{ "id": "d66dedb4-ed78-4b75-b7bf-4814fd4ab600", "revision": 0, "last_node_id": 73, "last_link_id": 101, "nodes": [ { "id": 6, "type": "GetNode", "pos": [ 2150, -2250 ], "size": [ 210, 40 ], "flags": { "collapsed": true }, "order": 0, "mode": 0, "inputs": [], "outputs": [ { "name": "INT", "type": "INT", "links": [ 73 ] } ], "title": "Get_Seed", "properties": { "Node name for S&R": "GetNode", "aux_id": "kijai/ComfyUI-KJNodes", "ue_properties": { "widget_ue_connectable": {}, "version": "7.8", "input_ue_unconnectable": {} } }, "widgets_values": [ "Seed" ], "widgets_values_named": { "Constant": "Seed" }, "color": "#1b4669", "bgcolor": "#29699c" }, { "id": 11, "type": "Fast Groups Bypasser (rgthree)", "pos": [ 2150, -2350 ], "size": [ 220, 60 ], "flags": {}, "order": 1, "mode": 0, "inputs": [], "outputs": [ { "label": "🤫", "name": "OPT_CONNECTION", "type": "*", "links": [] } ], "title": "Relay Bypasser", "properties": { "matchColors": "", "matchTitle": "#Relay Upscale Video", "showNav": false, "showAllGraphs": true, "sort": "position", "customSortAlphabet": "", "toggleRestriction": "default", "ue_properties": { "widget_ue_connectable": {}, "input_ue_unconnectable": {}, "version": "7.8" } }, "color": "#c3782d", "bgcolor": "#af6419" }, { "id": 12, "type": "GetNode", "pos": [ 2150, -2210 ], "size": [ 210, 50 ], "flags": { "collapsed": true }, "order": 2, "mode": 0, "inputs": [], "outputs": [ { "name": "IMAGE", "type": "IMAGE", "links": [ 72, 95 ] } ], "title": "Get_Org Video", "properties": { "Node name for S&R": "GetNode", "aux_id": "kijai/ComfyUI-KJNodes", "ue_properties": { "widget_ue_connectable": {}, "version": "7.8", "input_ue_unconnectable": {} } }, "widgets_values": [ "Org Video" ], "widgets_values_named": { "Constant": "Org Video" }, "color": "#2a363b", "bgcolor": "#3f5159" }, { "id": 13, "type": "LCLoadVideo", "pos": [ 1670, -2400 ], "size": [ 270, 330 ], "flags": {}, "order": 3, "mode": 0, "inputs": [], "outputs": [ { "name": "images", "type": "IMAGE", "links": [ 55 ] }, { "name": "audio", "type": "AUDIO", "links": [ 53 ] }, { "name": "fps", "type": "FLOAT", "links": [ 54 ] }, { "name": "frame_count", "type": "INT", "links": null } ], "properties": { "cnr_id": "ComfyUI_LC_AV_nodes", "ver": "23ab7e5f491f9d01fe37e024791bcd0bf3f0757b", "Node name for S&R": "LCLoadVideo", "ue_properties": { "widget_ue_connectable": {}, "input_ue_unconnectable": {}, "version": "7.8" } }, "widgets_values": [ "\"H:\\ComfyUI\\output\\Videos\\Minimax_2026-09-17_00006.mp4\"" ], "widgets_values_named": { "path": "\"H:\\ComfyUI\\output\\Videos\\Minimax_2026-09-17_00006.mp4\"" }, "color": "#1a5a7a", "bgcolor": "#1a5a7a" }, { "id": 15, "type": "LoadImage", "pos": [ 1670, -3110 ], "size": [ 350, 510 ], "flags": {}, "order": 4, "mode": 0, "inputs": [], "outputs": [ { "name": "IMAGE", "type": "IMAGE", "links": [ 52, 84 ] }, { "name": "MASK", "type": "MASK", "links": null } ], "properties": { "cnr_id": "comfy-core", "ver": "0.44.0", "Node name for S&R": "LoadImage", "ue_properties": { "widget_ue_connectable": {}, "input_ue_unconnectable": {}, "version": "7.8" } }, "widgets_values": [ "Lonecats Krea2 Pro_00040.png", "image" ], "widgets_values_named": { "image": "Lonecats Krea2 Pro_00040.png", "upload": "image" }, "color": "#233", "bgcolor": "#355" }, { "id": 20, "type": "LCSeed", "pos": [ 1200, -2960 ], "size": [ 270, 130 ], "flags": {}, "order": 5, "mode": 0, "inputs": [], "outputs": [ { "name": "seed", "type": "INT", "links": [ 17 ] } ], "properties": { "cnr_id": "ComfyUI_LC123_nodes", "ver": "e9d3f5fa5d7d6c0f603b2e19f8a8bf458ee3b8ec", "Node name for S&R": "LCSeed", "lc_seed_history": [ 8675309, 8675310, 8675309 ], "ue_properties": { "widget_ue_connectable": {}, "input_ue_unconnectable": {}, "version": "7.8" } }, "widgets_values": [ 8675309, "", "", "— last used seeds —" ], "widgets_values_named": { "base_seed": 8675309, "🎲 Randomize Each Time": "", "🎲 New Fixed Random": "", "seed history": "— last used seeds —" }, "color": "#822305", "bgcolor": "#822305" }, { "id": 21, "type": "SetNode", "pos": [ 1490, -2960 ], "size": [ 210, 60 ], "flags": { "collapsed": true }, "order": 27, "mode": 0, "inputs": [ { "name": "INT", "type": "INT", "link": 17 } ], "outputs": [ { "name": "INT", "type": "INT", "links": null } ], "title": "Set_Seed", "properties": { "Node name for S&R": "SetNode", "aux_id": "kijai/ComfyUI-KJNodes", "previousName": "Seed", "ue_properties": { "widget_ue_connectable": {}, "version": "7.8", "input_ue_unconnectable": {} } }, "widgets_values": [ "Seed" ], "widgets_values_named": { "Constant": "Seed" }, "color": "#1b4669", "bgcolor": "#29699c" }, { "id": 27, "type": "SetNode", "pos": [ 2040, -3110 ], "size": [ 210, 60 ], "flags": { "collapsed": true }, "order": 26, "mode": 0, "inputs": [ { "name": "IMAGE", "type": "IMAGE", "link": 52 } ], "outputs": [ { "name": "IMAGE", "type": "IMAGE", "links": null } ], "title": "Set_org image", "properties": { "Node name for S&R": "SetNode", "aux_id": "kijai/ComfyUI-KJNodes", "previousName": "org image", "ue_properties": { "widget_ue_connectable": {}, "version": "7.8", "input_ue_unconnectable": {} } }, "widgets_values": [ "org image" ], "widgets_values_named": { "Constant": "org image" }, "color": "#2a363b", "bgcolor": "#3f5159" }, { "id": 28, "type": "LCBatchImageComparer", "pos": [ 3080, -3070 ], "size": [ 380, 540 ], "flags": {}, "order": 36, "mode": 0, "inputs": [ { "label": "Before", "name": "image_a", "shape": 7, "type": "IMAGE", "link": 34 }, { "label": "After Relay", "name": "image_b", "shape": 7, "type": "IMAGE", "link": 71 } ], "outputs": [ { "name": "images", "type": "IMAGE", "links": [] } ], "properties": { "cnr_id": "ComfyUI_LC123_nodes", "ver": "e9d3f5fa5d7d6c0f603b2e19f8a8bf458ee3b8ec", "Node name for S&R": "LCBatchImageComparer", "comparer_mode": "Slide", "ue_properties": { "widget_ue_connectable": {}, "input_ue_unconnectable": {}, "version": "7.8" } }, "color": "#324B4B", "bgcolor": "#324B4B" }, { "id": 33, "type": "GetNode", "pos": [ 2210, -2970 ], "size": [ 210, 40 ], "flags": { "collapsed": true }, "order": 6, "mode": 0, "inputs": [], "outputs": [ { "name": "INT", "type": "INT", "links": [ 68 ] } ], "title": "Get_Seed", "properties": { "Node name for S&R": "GetNode", "aux_id": "kijai/ComfyUI-KJNodes", "ue_properties": { "widget_ue_connectable": {}, "version": "7.8", "input_ue_unconnectable": {} } }, "widgets_values": [ "Seed" ], "widgets_values_named": { "Constant": "Seed" }, "color": "#1b4669", "bgcolor": "#29699c" }, { "id": 36, "type": "SetNode", "pos": [ 2210, -2890 ], "size": [ 210, 50 ], "flags": { "collapsed": true }, "order": 35, "mode": 0, "inputs": [ { "name": "IMAGE", "type": "IMAGE", "link": 70 } ], "outputs": [ { "name": "IMAGE", "type": "IMAGE", "links": [] } ], "title": "Set_Relay Image", "properties": { "Node name for S&R": "SetNode", "aux_id": "kijai/ComfyUI-KJNodes", "previousName": "Relay Image", "ue_properties": { "widget_ue_connectable": {}, "version": "7.8", "input_ue_unconnectable": {} } }, "widgets_values": [ "Relay Image" ], "widgets_values_named": { "Constant": "Relay Image" }, "color": "#2a363b", "bgcolor": "#3f5159" }, { "id": 38, "type": "Fast Groups Bypasser (rgthree)", "pos": [ 2210, -3070 ], "size": [ 220, 60 ], "flags": {}, "order": 7, "mode": 0, "inputs": [], "outputs": [ { "label": "🤫", "name": "OPT_CONNECTION", "type": "*", "links": [] } ], "title": "Relay Bypasser", "properties": { "matchColors": "", "matchTitle": "#Relay Upscale Image", "showNav": false, "showAllGraphs": true, "sort": "position", "customSortAlphabet": "", "toggleRestriction": "default", "ue_properties": { "widget_ue_connectable": {}, "input_ue_unconnectable": {}, "version": "7.8" } }, "color": "#c3782d", "bgcolor": "#af6419" }, { "id": 39, "type": "GetNode", "pos": [ 2210, -2930 ], "size": [ 210, 50 ], "flags": { "collapsed": true }, "order": 8, "mode": 0, "inputs": [], "outputs": [ { "name": "IMAGE", "type": "IMAGE", "links": [ 34, 67, 99 ] } ], "title": "Get_org image", "properties": { "Node name for S&R": "GetNode", "aux_id": "kijai/ComfyUI-KJNodes", "ue_properties": { "widget_ue_connectable": {}, "version": "7.8", "input_ue_unconnectable": {} } }, "widgets_values": [ "org image" ], "widgets_values_named": { "Constant": "org image" }, "color": "#2a363b", "bgcolor": "#3f5159" }, { "id": 40, "type": "SetNode", "pos": [ 1960, -2350 ], "size": [ 210, 60 ], "flags": { "collapsed": true }, "order": 24, "mode": 0, "inputs": [ { "name": "AUDIO", "type": "AUDIO", "link": 53 } ], "outputs": [ { "name": "AUDIO", "type": "AUDIO", "links": null } ], "title": "Set_Audio", "properties": { "Node name for S&R": "SetNode", "aux_id": "kijai/ComfyUI-KJNodes", "previousName": "Audio", "ue_properties": { "widget_ue_connectable": {}, "version": "7.8", "input_ue_unconnectable": {} } }, "widgets_values": [ "Audio" ], "widgets_values_named": { "Constant": "Audio" } }, { "id": 41, "type": "SetNode", "pos": [ 1970, -2310 ], "size": [ 210, 60 ], "flags": { "collapsed": true }, "order": 25, "mode": 0, "inputs": [ { "name": "FLOAT", "type": "FLOAT", "link": 54 } ], "outputs": [ { "name": "FLOAT", "type": "FLOAT", "links": [] } ], "title": "Set_FPS", "properties": { "Node name for S&R": "SetNode", "aux_id": "kijai/ComfyUI-KJNodes", "previousName": "FPS", "ue_properties": { "widget_ue_connectable": {}, "version": "7.8", "input_ue_unconnectable": {} } }, "widgets_values": [ "FPS" ], "widgets_values_named": { "Constant": "FPS" }, "color": "#232", "bgcolor": "#353" }, { "id": 42, "type": "SetNode", "pos": [ 1970, -2390 ], "size": [ 210, 60 ], "flags": { "collapsed": true }, "order": 23, "mode": 0, "inputs": [ { "name": "IMAGE", "type": "IMAGE", "link": 55 } ], "outputs": [ { "name": "IMAGE", "type": "IMAGE", "links": null } ], "title": "Set_Org Video", "properties": { "Node name for S&R": "SetNode", "aux_id": "kijai/ComfyUI-KJNodes", "previousName": "Org Video", "ue_properties": { "widget_ue_connectable": {}, "version": "7.8", "input_ue_unconnectable": {} } }, "widgets_values": [ "Org Video" ], "widgets_values_named": { "Constant": "Org Video" }, "color": "#2a363b", "bgcolor": "#3f5159" }, { "id": 43, "type": "GetNode", "pos": [ 2150, -2170 ], "size": [ 210, 60 ], "flags": { "collapsed": true }, "order": 9, "mode": 0, "inputs": [], "outputs": [ { "name": "AUDIO", "type": "AUDIO", "links": [ 96 ] } ], "title": "Get_Audio", "properties": { "Node name for S&R": "GetNode", "aux_id": "kijai/ComfyUI-KJNodes", "ue_properties": { "widget_ue_connectable": {}, "version": "7.8", "input_ue_unconnectable": {} } }, "widgets_values": [ "Audio" ], "widgets_values_named": { "Constant": "Audio" }, "color": "#322", "bgcolor": "#533" }, { "id": 44, "type": "GetNode", "pos": [ 2150, -2130 ], "size": [ 210, 60 ], "flags": { "collapsed": true }, "order": 10, "mode": 0, "inputs": [], "outputs": [ { "name": "FLOAT", "type": "FLOAT", "links": [ 97 ] } ], "title": "Get_FPS", "properties": { "Node name for S&R": "GetNode", "aux_id": "kijai/ComfyUI-KJNodes", "ue_properties": { "widget_ue_connectable": {}, "version": "7.8", "input_ue_unconnectable": {} } }, "widgets_values": [ "FPS" ], "widgets_values_named": { "Constant": "FPS" }, "color": "#232", "bgcolor": "#353" }, { "id": 45, "type": "GetNode", "pos": [ 3480, -3100 ], "size": [ 210, 60 ], "flags": { "collapsed": true }, "order": 11, "mode": 0, "inputs": [], "outputs": [ { "name": "IMAGE", "type": "IMAGE", "links": [ 59 ] } ], "title": "Get_Relay Image", "properties": { "Node name for S&R": "GetNode", "aux_id": "kijai/ComfyUI-KJNodes", "ue_properties": { "widget_ue_connectable": {}, "version": "7.8", "input_ue_unconnectable": {} } }, "widgets_values": [ "Relay Image" ], "widgets_values_named": { "Constant": "Relay Image" }, "color": "#2a363b", "bgcolor": "#3f5159" }, { "id": 46, "type": "LCEasyFolder", "pos": [ 2730, -2350 ], "size": [ 300, 160 ], "flags": {}, "order": 12, "mode": 0, "inputs": [], "outputs": [ { "name": "filename_prefix", "type": "STRING", "links": [ 98 ] } ], "properties": { "cnr_id": "ComfyUI_LC123_nodes", "ver": "e9d3f5fa5d7d6c0f603b2e19f8a8bf458ee3b8ec", "Node name for S&R": "LCEasyFolder", "ue_properties": { "widget_ue_connectable": {}, "input_ue_unconnectable": {}, "version": "7.8" } }, "widgets_values": [ "Relay", "Video Test", "", "none", "auto" ], "widgets_values_named": { "folder": "Relay", "prefix": "Video Test", "suffix": "", "timestamp": "none", "path_separator": "auto" }, "color": "#324B4B", "bgcolor": "#324B4B" }, { "id": 47, "type": "LCEasyFolder", "pos": [ 3480, -3060 ], "size": [ 300, 160 ], "flags": {}, "order": 13, "mode": 0, "inputs": [], "outputs": [ { "name": "filename_prefix", "type": "STRING", "links": [ 60, 81 ] } ], "properties": { "cnr_id": "ComfyUI_LC123_nodes", "ver": "e9d3f5fa5d7d6c0f603b2e19f8a8bf458ee3b8ec", "Node name for S&R": "LCEasyFolder", "ue_properties": { "widget_ue_connectable": {}, "input_ue_unconnectable": {}, "version": "7.8" } }, "widgets_values": [ "Relay", "Image Test", "", "none", "auto" ], "widgets_values_named": { "folder": "Relay", "prefix": "Image Test", "suffix": "", "timestamp": "none", "path_separator": "auto" }, "color": "#324B4B", "bgcolor": "#324B4B" }, { "id": 48, "type": "SaveImage", "pos": [ 3790, -3100 ], "size": [ 390, 560 ], "flags": {}, "order": 29, "mode": 0, "inputs": [ { "name": "images", "type": "IMAGE", "link": 59 }, { "name": "filename_prefix", "type": "STRING", "widget": { "name": "filename_prefix" }, "link": 60 } ], "outputs": [ { "name": "images", "type": "IMAGE", "links": [ 85 ] } ], "properties": { "cnr_id": "comfy-core", "ver": "0.44.0", "Node name for S&R": "SaveImage", "ue_properties": { "widget_ue_connectable": {}, "input_ue_unconnectable": {}, "version": "7.8" } }, "widgets_values": [ "ComfyUI" ], "widgets_values_named": { "filename_prefix": "ComfyUI" }, "color": "#233", "bgcolor": "#355" }, { "id": 49, "type": "Fast Groups Bypasser (rgthree)", "pos": [ 1200, -3110 ], "size": [ 320, 100 ], "flags": {}, "order": 14, "mode": 0, "inputs": [], "outputs": [ { "name": "OPT_CONNECTION", "type": "*", "links": null } ], "properties": { "matchColors": "", "matchTitle": ":", "showNav": false, "showAllGraphs": true, "sort": "position", "customSortAlphabet": "", "toggleRestriction": "always one", "ue_properties": { "widget_ue_connectable": {}, "input_ue_unconnectable": {}, "version": "7.8" } }, "color": "#c87a2c", "bgcolor": "#b46618" }, { "id": 50, "type": "LCTimer", "pos": [ 1200, -2780 ], "size": [ 300, 250 ], "flags": {}, "order": 15, "mode": 0, "inputs": [], "outputs": [], "properties": { "cnr_id": "ComfyUI_LC123_nodes", "ver": "e9d3f5fa5d7d6c0f603b2e19f8a8bf458ee3b8ec", "Node name for S&R": "LCTimer", "ue_properties": { "widget_ue_connectable": {}, "input_ue_unconnectable": {}, "version": "7.8" }, "lc_runs": [ { "t": 12.64, "at": "15:44", "ok": true }, { "t": 1.3, "at": "15:46", "ok": true }, { "t": 34.39, "at": "15:47", "ok": false }, { "t": 0.8, "at": "15:49", "ok": true }, { "t": 5.27, "at": "15:50", "ok": true }, { "t": 5.37, "at": "15:50", "ok": true }, { "t": 2.05, "at": "15:52", "ok": true }, { "t": 2.92, "at": "15:52", "ok": true }, { "t": 73.64, "at": "16:12", "ok": true }, { "t": 0.78, "at": "16:13", "ok": true }, { "t": 74.34, "at": "18:52", "ok": true }, { "t": 74.72, "at": "18:59", "ok": true }, { "t": 2.85, "at": "19:03", "ok": true }, { "t": 16.47, "at": "19:39", "ok": false }, { "t": 49.13, "at": "19:42", "ok": true }, { "t": 2.93, "at": "19:42", "ok": true }, { "t": 6.93, "at": "18:17", "ok": true }, { "t": 0.89, "at": "18:18", "ok": true }, { "t": 0.79, "at": "18:19", "ok": true }, { "t": 7.65, "at": "18:19", "ok": true } ] }, "widgets_values": [ "" ], "widgets_values_named": { "lc_rt": "" }, "color": "#000000", "bgcolor": "#000000" }, { "id": 52, "type": "LCLabel", "pos": [ 1800, -3340 ], "size": [ 1159, 131 ], "flags": {}, "order": 16, "mode": 0, "inputs": [], "outputs": [], "properties": { "text": "Relay VSR Upscale Module\n", "font": "Oswald", "size": 107, "bold": true, "italic": false, "color": "#ffffff", "align": "center", "outline": false, "outlineColor": "#000000", "outlineWidth": 4, "shadow": false, "shadowBlur": 10, "bg": false, "bgColor": "#324B4B", "radius": 12, "padding": 8, "angle": 0, "width": 0, "cnr_id": "ComfyUI_LC123_nodes", "ver": "e9d3f5fa5d7d6c0f603b2e19f8a8bf458ee3b8ec", "Node name for S&R": "LCLabel", "_w": 1159, "_h": 131, "ue_properties": { "widget_ue_connectable": {}, "input_ue_unconnectable": {}, "version": "7.8" } }, "color": "#fff0", "bgcolor": "#fff0" }, { "id": 57, "type": "LCRelayVSRUpscale", "pos": [ 2760, -3030 ], "size": [ 300, 250 ], "flags": {}, "order": 32, "mode": 0, "inputs": [ { "name": "image", "type": "IMAGE", "link": 67 }, { "name": "protect_mask", "shape": 7, "type": "MASK", "link": 100 }, { "name": "seed", "type": "INT", "widget": { "name": "seed" }, "link": 68 } ], "outputs": [ { "name": "image", "type": "IMAGE", "links": [ 70, 71 ] }, { "name": "relay", "type": "LC_RELAY", "links": [] }, { "name": "protected", "type": "MASK", "links": null } ], "properties": { "cnr_id": "ComfyUI_LC_MaskMaker_nodes", "ver": "0.29.8", "Node name for S&R": "LCRelayVSRUpscale", "ue_properties": { "widget_ue_connectable": { "seed": true }, "version": "7.8", "input_ue_unconnectable": {} } }, "widgets_values": [ "image", "2x", 8675309, "fixed", 1, 0.5, true ], "widgets_values_named": { "source": "image", "upscale": "2x", "seed": 8675309, "control_after_generate": "fixed", "detail": 1, "mask_protect_strength": 0.5, "detect_faces": true }, "color": "#324b4b", "bgcolor": "#324b4b" }, { "id": 58, "type": "LCRelayVSRUpscale", "pos": [ 2390, -2310 ], "size": [ 320, 250 ], "flags": {}, "order": 22, "mode": 0, "inputs": [ { "name": "image", "type": "IMAGE", "link": 72 }, { "name": "protect_mask", "shape": 7, "type": "MASK", "link": null }, { "name": "seed", "type": "INT", "widget": { "name": "seed" }, "link": 73 } ], "outputs": [ { "name": "image", "type": "IMAGE", "links": [] }, { "name": "relay", "type": "LC_RELAY", "links": [ 94 ] }, { "name": "protected", "type": "MASK", "links": null } ], "properties": { "cnr_id": "ComfyUI_LC_MaskMaker_nodes", "ver": "0.29.8", "Node name for S&R": "LCRelayVSRUpscale", "ue_properties": { "widget_ue_connectable": { "seed": true }, "version": "7.8", "input_ue_unconnectable": {} } }, "widgets_values": [ "video", "2x", 8675309, "fixed", 1, 0.5, true ], "widgets_values_named": { "source": "video", "upscale": "2x", "seed": 8675309, "control_after_generate": "fixed", "detail": 1, "mask_protect_strength": 0.5, "detect_faces": true }, "color": "#324b4b", "bgcolor": "#324b4b" }, { "id": 59, "type": "LCLabel", "pos": [ 3000, -3330 ], "size": [ 262, 121 ], "flags": {}, "order": 17, "mode": 0, "inputs": [], "outputs": [], "properties": { "text": "V3.0.2", "font": "Oswald", "size": 98, "bold": true, "italic": false, "color": "#4da3ff", "align": "left", "outline": false, "outlineColor": "#000000", "outlineWidth": 4, "shadow": false, "shadowBlur": 10, "bg": false, "bgColor": "#324B4B", "radius": 12, "padding": 8, "angle": 0, "width": 0, "cnr_id": "ComfyUI_LC123_nodes", "ver": "e9d3f5fa5d7d6c0f603b2e19f8a8bf458ee3b8ec", "Node name for S&R": "LCLabel", "_w": 262, "_h": 121, "ue_properties": { "widget_ue_connectable": {}, "input_ue_unconnectable": {}, "version": "7.8" } }, "color": "#fff0", "bgcolor": "#fff0" }, { "id": 61, "type": "LCTextOverlay", "pos": [ 4560, -3100 ], "size": [ 300, 640 ], "flags": {}, "order": 37, "mode": 0, "inputs": [ { "name": "image", "type": "IMAGE", "link": 86 } ], "outputs": [ { "name": "image", "type": "IMAGE", "links": [ 79 ] } ], "properties": { "cnr_id": "ComfyUI_LC123_nodes", "ver": "e9d3f5fa5d7d6c0f603b2e19f8a8bf458ee3b8ec", "Node name for S&R": "LCTextOverlay", "ue_properties": { "widget_ue_connectable": {}, "input_ue_unconnectable": {}, "version": "7.8" }, "lc_overlay_text": "Before After", "lc_preview_meta": [ { "filename": "lc_before_temp_kghcp_00002_.png", "subfolder": "", "type": "temp", "metadata": { "kind": "image", "width": 2144, "height": 2704 }, "id": "30f22e23-b89e-47af-a481-00a42be6c40b" } ] }, "widgets_values": [ "Before After", "Arial", 110, 255, 255, 255, 50.3, 85.5, "center" ], "widgets_values_named": { "text": "Before After", "font": "Arial", "font_size": 110, "color_r": 255, "color_g": 255, "color_b": 255, "x_percent": 50.3, "y_percent": 85.5, "alignment": "center" }, "color": "#324B4B", "bgcolor": "#324B4B" }, { "id": 62, "type": "SaveImage", "pos": [ 4880, -3010 ], "size": [ 230, 350 ], "flags": {}, "order": 38, "mode": 0, "inputs": [ { "name": "images", "type": "IMAGE", "link": 79 }, { "name": "filename_prefix", "type": "STRING", "widget": { "name": "filename_prefix" }, "link": 82 } ], "outputs": [ { "name": "images", "type": "IMAGE", "links": null } ], "properties": { "cnr_id": "comfy-core", "ver": "0.44.0", "Node name for S&R": "SaveImage", "ue_properties": { "widget_ue_connectable": {}, "input_ue_unconnectable": {}, "version": "7.8" } }, "widgets_values": [ "ComfyUI" ], "widgets_values_named": { "filename_prefix": "ComfyUI" }, "color": "#233", "bgcolor": "#355" }, { "id": 63, "type": "PrimitiveString", "pos": [ 4870, -3060 ], "size": [ 270, 60 ], "flags": { "collapsed": true }, "order": 18, "mode": 0, "inputs": [], "outputs": [ { "name": "STRING", "type": "STRING", "links": [ 80 ] } ], "properties": { "cnr_id": "comfy-core", "ver": "0.44.0", "Node name for S&R": "PrimitiveString", "ue_properties": { "widget_ue_connectable": {}, "input_ue_unconnectable": {}, "version": "7.8" } }, "widgets_values": [ "Stitch" ], "widgets_values_named": { "value": "Stitch" } }, { "id": 64, "type": "LCJoinStrings", "pos": [ 4870, -3100 ], "size": [ 330, 130 ], "flags": { "collapsed": true }, "order": 30, "mode": 0, "inputs": [ { "name": "string_1", "shape": 7, "type": "STRING", "link": 80 }, { "name": "string_2", "shape": 7, "type": "STRING", "link": 81 } ], "outputs": [ { "name": "string", "type": "STRING", "links": [ 82 ] } ], "properties": { "cnr_id": "ComfyUI_LC123_nodes", "ver": "e9d3f5fa5d7d6c0f603b2e19f8a8bf458ee3b8ec", "Node name for S&R": "LCJoinStrings", "ue_properties": { "widget_ue_connectable": {}, "input_ue_unconnectable": {}, "version": "7.8" } }, "widgets_values": [ 2, " " ], "widgets_values_named": { "inputcount": 2, "delimiter": " " }, "color": "#28281E", "bgcolor": "#28281E" }, { "id": 65, "type": "LCImageSplit", "pos": [ 4250, -3100 ], "size": [ 300, 500 ], "flags": {}, "order": 34, "mode": 0, "inputs": [ { "name": "image_a", "type": "IMAGE", "link": 84 }, { "name": "image_b", "type": "IMAGE", "link": 85 } ], "outputs": [ { "name": "split 🖼️", "type": "IMAGE", "links": [ 86 ] } ], "properties": { "cnr_id": "ComfyUI_LC123_nodes", "ver": "e9d3f5fa5d7d6c0f603b2e19f8a8bf458ee3b8ec", "Node name for S&R": "LCImageSplit", "ue_properties": { "widget_ue_connectable": {}, "input_ue_unconnectable": {}, "version": "7.8" } }, "widgets_values": [ 0.5, true, 5, "#FFFFFF" ], "widgets_values_named": { "split_position": 0.5, "show_divider": true, "divider_width": 5, "divider_color": "#FFFFFF" }, "color": "#324B4B", "bgcolor": "#324B4B" }, { "id": 68, "type": "LCNote", "pos": [ 2730, -2150 ], "size": [ 300, 400 ], "flags": {}, "order": 19, "mode": 0, "inputs": [], "outputs": [], "properties": { "lc_note": { "source": "en", "texts": { "en": { "text": "## Save💾\n\n**LC Save Video**\n\n- **relay**: Takes the output and saves the video a few frames at a time, straight to the file. The whole upscaled video never sits in RAM.\n - 5 s clip at 4x: **0.3 GB** of RAM instead of **12 GB**.\n- Leave the Relay node's **image** output empty, or the full video gets built anyway.\n\n\n\n**When bypassing**\n- bypass the Relay group and LC Save Video saves the original video from **images** instead. Same save node, nothing to rewire.\n\n- **save_metadata** puts the workflow in the mp4, so CivitAI can find your resources.\n\n**Settings ⚙️**\n\n- **crf:** 19 looks like the source. Lower = better and bigger.\n- **pingpong** and **loop_count** are skipped for a relay video.\n\n" }, "zh": { "text": "## 保存💾\n\n**LC 保存视频**\n\n- **relay**：逐帧将输出直接保存到文件，整个超分辨率视频不会占用内存。\n - 5秒4倍速视频：仅需 **0.3 GB** 内存，而非 **12 GB**。\n- 若 Relay 节点的 **image** 输出留空，则仍会完整生成整段视频。\n\n---\n\n**当绕过时**\n- 绕过 Relay 组后，LC 保存视频将从 **images** 直接保存原始视频。同一保存节点，无需重新布线。\n\n- **save_metadata** 将工作流嵌入 mp4 文件，以便 CivitAI 能识别你的资源。\n\n---\n\n**设置 ⚙️**\n\n- **crf**：19 看起来与源视频一致。数值越低，画质越好但文件越大。\n- 对于 relay 视频，**pingpong** 和 **loop_count** 设置将被跳过。", "src": "e76fffde", "srcTitle": "" }, "zh-TW": { "text": "## 儲存💾\n\n**LC 儲存影片**\n\n- **relay**：將輸出逐幀儲存至檔案，不讓整個上採樣後的影片停留在記憶體中。\n - 5 秒片段以 4x 上採樣：僅需 **0.3 GB** 記憶體，而非 **12 GB**。\n- 若 Relay 節點的 **image** 輸出留空，則仍會完整建立影片。\n\n---\n\n**當繞過時**\n- 繞過 Relay 群組後，LC 儲存影片將直接從 **images** 儲存原始影片。同一儲存節點，無需重新接線。\n\n- **save_metadata** 會將工作流程嵌入 mp4 檔案，讓 CivitAI 可搜尋到你的資源。\n\n---\n\n**設定 ⚙️**\n\n- **crf**：19 看起來與原始影片相近。數值越低 = 質量越好、檔案越大。\n- 對於 relay 影片，**pingpong** 和 **loop_count** 設定將被跳過。", "src": "e76fffde", "srcTitle": "" }, "ja": { "text": "## 保存💾\n\n**LC Save Video**\n\n- **relay**: 出力を受け取り、フレームごとにファイルに直接書き込み。アップスケールされた動画全体がRAMに一時的に残らない。\n - 4xで5秒のクリップ：**0.3 GB** のRAM（通常は**12 GB**）。\n- リレー・ノードの **image** 出力を空にしないと、動画全体が無条件に作成される。\n\n---\n\n**バイトパス時**\n- リレー・グループをバイパスすると、LC Save Video は **images** から元の動画を保存する。同じ保存ノードで、配線を変更する必要はない。\n\n- **save_metadata** はワークフロー情報をMP4に埋め込み、CivitAIがリソースを検索できるようにする。\n\n---\n\n**設定 ⚙️**\n\n- **crf:** 19は元の動画とほぼ同じ品質。低い値ほど品質は高くなりますが、ファイルサイズも大きくなる。\n- リレー・ビデオでは **pingpong** と **loop_count** は無視される。", "src": "e76fffde", "srcTitle": "" }, "ko": { "text": "## 저장💾\n\n**LC Save Video**\n\n- **relay**: 출력을 받아 몇 프레임씩 직접 파일에 저장합니다. 확대된 전체 영상은 RAM에 한 번도 저장되지 않습니다.\n - 4x 확대 5초 영상: **0.3 GB**의 RAM 대신 **12 GB** 사용.\n- Relay 노드의 **image** 출력을 빈 값으로 두어야만 전체 영상이 생성되지 않습니다.\n\n---\n\n**바이패스 시**\n- Relay 그룹을 우회하면 LC Save Video는 **images**에서 원본 영상을 저장합니다. 동일한 저장 노드이며, 재연결할 필요 없음.\n\n- **save_metadata**: 작업 흐름을 mp4에 포함시켜 CivitAI가 자원을 찾을 수 있도록 합니다.\n\n---\n\n**설정 ⚙️**\n\n- **crf:** 19는 원본과 유사한 품질입니다. 낮을수록 품질은 좋고 파일 크기는 커집니다.\n- **pingpong** 및 **loop_count**는 릴레이 영상에서는 무시됩니다.", "src": "e76fffde", "srcTitle": "" } }, "titles": {} }, "ue_properties": { "widget_ue_connectable": {}, "version": "7.8", "input_ue_unconnectable": {} } }, "widgets_values": [ "English ✏️", "## Save💾\n\n**LC Save Video**\n\n- **relay**: Takes the output and saves the video a few frames at a time, straight to the file. The whole upscaled video never sits in RAM.\n - 5 s clip at 4x: **0.3 GB** of RAM instead of **12 GB**.\n- Leave the Relay node's **image** output empty, or the full video gets built anyway.\n\n\n\n**When bypassing**\n- bypass the Relay group and LC Save Video saves the original video from **images** instead. Same save node, nothing to rewire.\n\n- **save_metadata** puts the workflow in the mp4, so CivitAI can find your resources.\n\n**Settings ⚙️**\n\n- **crf:** 19 looks like the source. Lower = better and bigger.\n- **pingpong** and **loop_count** are skipped for a relay video.\n\n" ], "widgets_values_named": { "language": "English ✏️", "text": "## Save💾\n\n**LC Save Video**\n\n- **relay**: Takes the output and saves the video a few frames at a time, straight to the file. The whole upscaled video never sits in RAM.\n - 5 s clip at 4x: **0.3 GB** of RAM instead of **12 GB**.\n- Leave the Relay node's **image** output empty, or the full video gets built anyway.\n\n\n\n**When bypassing**\n- bypass the Relay group and LC Save Video saves the original video from **images** instead. Same save node, nothing to rewire.\n\n- **save_metadata** puts the workflow in the mp4, so CivitAI can find your resources.\n\n**Settings ⚙️**\n\n- **crf:** 19 looks like the source. Lower = better and bigger.\n- **pingpong** and **loop_count** are skipped for a relay video.\n\n" }, "color": "#2E3A46", "bgcolor": "#2E3A46" }, { "id": 69, "type": "LCSaveVideo", "pos": [ 3040, -2350 ], "size": [ 300, 780 ], "flags": {}, "order": 31, "mode": 0, "inputs": [ { "name": "images", "shape": 7, "type": "IMAGE", "link": 95 }, { "name": "audio", "shape": 7, "type": "AUDIO", "link": 96 }, { "name": "av_pipe", "shape": 7, "type": "LC_AV_PIPE", "link": null }, { "name": "metadata", "shape": 7, "type": "LC_SAVE_META,LC_SAVE_VIDEO_META", "link": null }, { "name": "relay", "shape": 7, "type": "LC_RELAY", "link": 94 }, { "name": "frame_rate", "type": "FLOAT", "widget": { "name": "frame_rate" }, "link": 97 }, { "name": "filename_prefix", "type": "STRING", "widget": { "name": "filename_prefix" }, "link": 98 } ], "outputs": [ { "name": "saved_as", "type": "STRING", "links": [] } ], "properties": { "cnr_id": "ComfyUI_LC_AV_nodes", "ver": "0.5.21", "Node name for S&R": "LCSaveVideo", "ue_properties": { "widget_ue_connectable": { "frame_rate": true, "loop_count": true, "filename_prefix": true, "format": true, "pix_fmt": true, "crf": true, "save_metadata": true, "civitai_still": true, "pingpong": true, "save_output": true, "prune_outputs": true }, "version": "7.8", "input_ue_unconnectable": {} } }, "widgets_values": [ 24, 0, "Relay", "video/h264-mp4", "yuv420p", 19, false, false, false, true, true ], "widgets_values_named": { "frame_rate": 24, "loop_count": 0, "filename_prefix": "Relay", "format": "video/h264-mp4", "pix_fmt": "yuv420p", "crf": 19, "save_metadata": false, "civitai_still": false, "pingpong": false, "save_output": true, "prune_outputs": true }, "color": "#1a5a7a", "bgcolor": "#1a5a7a" }, { "id": 70, "type": "LCImageLabel", "pos": [ 1530, -3410 ], "size": [ 221, 221 ], "flags": {}, "order": 20, "mode": 0, "inputs": [], "outputs": [], "properties": { "imgW": 211, "imgH": 211, "angle": 0, "size": 211, "padding": 5, "borderWidth": 0, "borderColor": "#ffffff", "borderRadius": 0, "backgroundColor": "transparent", "bgTransparent": true, "syncImageRadius": true, "imageRadius": 0, "embeddedData": "data:image/webp;base64,UklGRrI6AABXRUJQVlA4WAoAAAAwAAAA/wAA/wAASUNDUMgBAAAAAAHIAAAAAAQwAABtbnRyUkdCIFhZWiAH4AABAAEAAAAAAABhY3NwAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAQAA9tYAAQAAAADTLQAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAlkZXNjAAAA8AAAACRyWFlaAAABFAAAABRnWFlaAAABKAAAABRiWFlaAAABPAAAABR3dHB0AAABUAAAABRyVFJDAAABZAAAAChnVFJDAAABZAAAAChiVFJDAAABZAAAAChjcHJ0AAABjAAAADxtbHVjAAAAAAAAAAEAAAAMZW5VUwAAAAgAAAAcAHMAUgBHAEJYWVogAAAAAAAAb6IAADj1AAADkFhZWiAAAAAAAABimQAAt4UAABjaWFlaIAAAAAAAACSgAAAPhAAAts9YWVogAAAAAAAA9tYAAQAAAADTLXBhcmEAAAAAAAQAAAACZmYAAPKnAAANWQAAE9AAAApbAAAAAAAAAABtbHVjAAAAAAAAAAEAAAAMZW5VUwAAACAAAAAcAEcAbwBvAGcAbABlACAASQBuAGMALgAgADIAMAAxADZBTFBI2BYAAAHAh/1/Grm1frI9PMubLIQZNmmgzMxNe8twe3KgzMx0+JSZmU+bcnvCpTBDs2lgs+HsbpZ3dnbAY+k3f1iW5RnPzOUbERMA/+///7mbFHpV5xZ4yqefksKupHuOUtjdwpYXdt5fcZta0B0SxVatkFN+YNjlLeQGRhB7wgUceZEi9vQr4Ir3ImLH4ALuGh0RGyoKN18jIrK3lMLtBB0Ro+OhYFdXISIu9hZuU2OISC+Hgl2ZzxCxt6pwq44gIq7UCjbyKEVEehkU7MFdiIi9/Qq305KmH7WCTVmCiGgcDwX7oD5TNCji8Xs0pXAi7zLTWo8VOaCxfefaL+46rDzkU0kBVNSCiEjfHeD3l1SPPeL035zzFUNEZEayd+eCB+tCXqWgIf6LUiaW0A3GEBnSngiKUr27/rXDQmphooZqz/7npgjaZkzITOP7vp1WohQYin/Cw8s7U4xhpqZaPjk6QAoFopVd9sXuOMUMZ/HNTw/WCgG1+rY5bSmGWUk7Z58UyPNI6LSvW1OYzfFlvy8i+Zs24vktMYbZnlhxfZjkZ4HTv+8y0BVjC64qzcOClyyOMnTNvp+O8eZZgWkLY+iuXf8cRPIo77Sf+9B12fYHwvmSOvVfPejKydkHKPkQGfJGC0OXZnvuCuY/was2p9DFE98MzHPIsK+i6O6s/mxPPuP/QyND129/oiJ/GfZlHLPW6OmOJGlmYPLbQXmK9m9bGGZv1xBNC4363Sd749Q5pAsPIflI2QtdmM29AeAqpSfNaEs5hWzD6Ur+MWxmErM6VccDAOI/7J026gzizsu1fGPiMorZzf5CrACAhKet0p3B5uu8eYVyYQNmOKPJ9qYmKoIrvUIAoAx7r88RbLvfl0d47tyPGcwSneufOavaqyglcaGuGjsApObdhBPYfZ8vbwg90YMZ3PrmQK8CXJ8ulLrZHgAZ/r3hAPY84MsTyt+MYQanfucF6wOpEK4MSABQj9nlAEZu9+QFNd8kMJO7y8GafIXiPeOkAIT/lJSHbZcqeUDV1wZmIrNa6RcItNlITiNygExolIc7zyA5r3KWgRloNG2mPHajpljVJWywGSGRoqAAQPgTJg03Hpbrir8y0PnYOwdULUZ+ou6EuR4eeZnawP3HVViR5/eeqQiA+oeINFwyPLeF3tfR8dgTw1So6rNofb3za4Xn3412jTV/9VnA6GTy20oBgDH7pNGPS3OZ99k4Ot379DgVQHmRWSRSPYOBPzJmCxObTiAW2iZkOw8lAtC/QRb2PeLNXcodEXTY+GqSBwCgogMFP9Z4Nd8xeyw+u8oCHkDE+B89AlBSLwtbzie5ipyxHx3u+U0xmMmfDYHkkYRT/OcYSkxFPqq0GGYgIv2uRADKNsnC1aNy1dE70Fm2aKIC3JIWFGwvAXPgzy1MhpHs+YvK87Si+ddaAajYKYt+Vpybhq5DZxN3FoPl+SmRjR5O0d9bUSZtXVYHlm9wcPdYARjULgljtyi5KPQddWb/MRpYeteg6BsKh4zeKufzEWB9Ig+bxgrA5KgkbByTg5T7EugkmzscBMfGRIzzCAcqO6QYDysCgygPdw8TgIupJPZhIPcc3oZO0hdLQFD9CUVjE4BLDtelsPdUgVDCArf1F1DeloSRi0muqapHJ41HQiA6sk+op5bn/YJJwc0BAa3VCjcWW0GgURIuGZxjPB9QJ5LXekBU+ZAJtZfywvtRbu9oAbJEAGd6rWCCLin5dy23nNWHDvaerYJwdSsK7w/zBvZJSaNxM7GCJ0Too4oVPCAJ90zMKUO2oYM9JyggrDxriDUHOcpfqBREnO8VuEwEE9MEfI2S2Hv+HOL9iDoQO4mA+NAuFG8KcHybUHZrlcDRQthaYwVTDDnYdmwOObEH5esXKiDufYfaaPZzSiOS0pi8XOBwMfzeY0XmSGIf+XNG2XqUT+/SwObkJrTZ4uMcrktCZO8rVqfa0A+yghEpObj/0FxBHtLlsbf8YFN9V7fT7jepH6Els4UNQavrbNBPVCtlviT6jjdHDGlG+d+WgN3wBrQbCZn8TVZt9mJjrL4VY1/7QXCsIQebDskN6rtU3q5qsD0iYiteYSqNmdKI+qOGjTTSmwmPNIt95QNRZZMk400tJ4xoR+l9B4PtwAPUll5nupAiYjqNuOukXhuIONfD05JCi3wgfhaTg01Dc4H6CZXGblJsqXc3o216KQFQfkZOGp86LmqvpYQXpCKbQmD2jrMIRyTpt5McML4TpS8Igu2K5cwevqYA+Dt5yM65hdlLHMcbzzBt0d0PuP/qUHnkA0m4tNj91PeZtI6RYH9cC1rrVuu8AGN1DqJ+6R60m0b2CuG8iphOc/QjgOtJ0gE8mEgl9ZzifmM6UHbqfGKPXJ6worusOkuA/AMt9i1lEnCHx6R2IKaRezfwKygutPD1SqJva26nvkel/RgA+955aL31ipRF6kjw7rBq1FFGYohpIMU0b65qcQViKsyDeZJwV7XbDWxD2ZGJIHFcxMq4IrTAAj9Rq5IcxNRCJiGN7FbT84hpNO8tAr6yDxFvsDhVVvwylyN/NmSxOzQJnhnMant/qOux2FF+DTOlEbcel5KCqxUAXwT5yQlgOSiFiI0Krx+VRN9W3S2wAWW3DgaJte1o/Q8VtOcZr3f6BrR4ubxdhDGrngDASYzHbgVL8iEiYqqK5+mThBsDAKCWVBOXOjImi92kWJFJIyuKS0MA6u2GVWIKAAxs5rF91IKe6n9DJLYSMW3G1FHgaUX+ItVqXMKE9/LIGlmRI8P9Tn6/t+1wd1K/YbJaK8G6aHci3te7WIPQRrTeFgIAcpXBQYqWff1goiGwbrIV/kubjvyuMrBUlyN3t8KBP8mi639u0xHZS4orlexHyewOIjCgDxGxwQ+HxKyMmwgAgG82T/AHDwS7rNgt4V6r7mMSvNSRYH0R5SVLeMfLsl7tdaXfp2R1VIHguSnTHJ8yk1l1VgN3SKsNdjEBZYFVYqD2DC+NLIr8p4jVgE7ksxN5Q5lDbSVupP6Ist9UBMg/KCKyz70VUbRe6OOR6YZYciAA3Gi11wuHMuRZ7/aDpfYTWn/Jq6QOxSe5UVmbrORkEPTMYYhInwi8wazofRoP/D+KNQQAYBK1eEeBct1OaixYT6cCbSqnlDlk3O5GVxmy6r0igZWchyojKDC71gJGtwv9RQGAyiSPXgigbeSlLR4D6yl9aJlGvZQTpg7hd4r7qMtRMr0ARMPbEBHZ41dQAaT1Ay3UO3QBOgoAwN/J0wcCwO8Qhfb4rOo6UIQewgk5tsnrPiVdsjrLhIqaTPT9/Si+digPJiUE+kImZQMvGgSAMobpNFrGJoHl5A4UfzxDOoPuc7oha5Yq1mYydqHdX0ZzlLeYQGs/E3mL16EBgKZjGq2XaxaH9aDNNZxyx2LlrkNmoGTjXBAOdXAMW9h4FAGA/hEUpF8PIwBwNW+3AgDhKIrW89Q/9KHdDs4wx5LDXEfbLquzv1hgPyIyAyU2X6ABXMdEkHU+EgA4nLcAgNS+T4WiRabijwy0nSCmyY7pR7pOUVTWTE3M32hCqb1/Ki/ajTZp/aHeg3iLfJVvxVCcXQigHL0LJRqq6WTmlHGV6xxtSKJXgbhvEUPpqSX/YHYQE/UzeV3z+tD2vsqaD3SUGSOmy9Fp9hhxm1eYpGSdDc/7VB6yFGZ4ZxIlJhv+ORXMzzqGMxSXUdaj5OYiG8rNugNumJh3gKqCmfzi3A+qy3haZX2j2oBJMRejKycoYKn1OLdCc5myPknsJmKnqNlG0k0i13hBMJR0bpPXZaakJBkHg13teyZE39/OXGNFDQhPos7tcZtrqaS+Glue+WJtFZUf6+7AZoRA/D10fr/LkA9QckvIVl0EhT9Wwf8X3Q3YK14Q93RlQLvLqGtkrdLsFK9AYeMsAPA9abjApxrYHGdkQIfPXTzNst4jNsi9ulikBgAg+AXLuoZisKksxgxs97qLLyrrbrA5oQPFNwVMMKAx2/TjwO4YPRPaXCaoS2Ln2qhdhzbfUjlkWizLZql2yFzMxGaX6WdIolPFimYxG/QSwoHgD9kVHQZ2D0xlxHaPu4yikowhQsF3DbSZnAB89XmWVfNUO76tmJGrNXc5WFayn4j/JR3t9va3KFqH2awfA3bvYJnxneouZzFJiVKB0Cs62t4bsjgkkVWdYTvj+zAj2fOKu1wpK15iVfutgfaX+XjaDJZVuzw2ShswM+mNxF0elJUo4mmH1jOU+KHGq2pDRJY9O7xinlmYoanjwVXJo7KSHFL+SifKZH9SOGQ6RYwb2dNdLKQ8STMlMcplnkZZYdPBexhKpVcRTslWTK1cy7LHOFlE/VsKMzVWkyOKTRMNlKtfpXBO6F4z/dk4Zi/7SLEKf2RgxvYUucyjTFa5qahXUurBsEk57LTA+R2YzRv8FmPWMszcPT6XuVeWXm3Slsp69mgNAMCjHr8Fs3pfDa9fB2byYtVd4PfShpvgN4ak28+e5DWdtR2zu/s83i2Y0S8RlzmVSjIO5PTvlXT1MaeeOSpQevVezHK6oMakNmQUvQxcdoIhiZ7L0RZJuurAYRfdes+MKGZ9auZEBeAMI6Pi49ymRha7jwMXG3KuPerIu9a2M3RBtvPOqhF7MKNbwm5TIgvfUjglzZJO25Ni6JJGpI1m1jzVbcK6rAUeDnmUyTCuPTWFLsowo9ltxG2CfbJ2BDgwoFcGvXNYg5tkuD4V3NbXKqu7kqe8xiTgXyvvY7mqrcx11F9kxcfwYFCvjM8qRsZy1Tce1yHvyjLOtFDeYBIaR5Z8znJT6lJw32uYJHzIAgZ1SohO8E7pzE1dg1zoIEPWv1QL8idqL3lqcdW7LCd973WhioSsPaUWEF5vj14bUMZ15CLjEnBhX7usvgOs4OCYLXxVA+9bLAd1VbuRukoWvVpAeYzaqvcDDNyfg+Z63Ig8ySTh15oVBBba6hoMoDxo5JzkmeDKRxiymgcIwJAWO8ZNCkBwYc7ZVuxOZXFZ+jkicFrcBq4PAsCw5hzDHlHcybNLFr6oiSh/NmxEJwEAuTKRWyKDwZ3JP6XtrBUBzwdMjL2kAYD2EnUd5gT7QnMpOMuQlbhQCAIzmRC2DAYACM51G4ZORg8Ft67okYWfBoSgeIEY/bMGAFC10V0YOvqt17W0ldKapohB2XIhbBltgpGNbsLQ0ehR4N73UlnGnaoYVK4Wws+9Jpja5B4Mnf3W62JDorJw4wAbULFKKHkxMcHh+9yCOdR7NLi4Z6W05FWKDShfyASwZTgHDtrhDgydZR9pbgY3MVm4foAdKPrIEMAlIQ4MXcJcgKHDnWPB1Wui0vQbPXbAe3efAPuHxoGSj/Xsc5r+SXE3dYU0bJhgC8gxe60wOZ1wQPt9u8ttLAWXv8CQlvpzwBZA1c/MAiNnEQ7AoMXUzRIXgtuXNknDllMkgPfmiAV2nKPwwH9Xq3uxzz2uR15h0tjM/hIAhsw3eBi5UuMBGTUv6VZ7BoL7j4tKw9iNmgzQzm3jYfLRMA/Ac+IK3ZX0yyEHarPl4Y5jpQCUv5HgIP1psAWA74JlSfdhr2q5AE6My2OLBskBMn6+bkJsPlm1APCetCjGXGZVCeTE4Ap5mHohKAdAOeRH3YSJ9/tZAWhTZnQYbtJ1JOTIS1PyMHKzKglAPWJ+AhGRdVzhtwIgwdO+69DdQr9VyRVFmx3AltOILABlxMvtBiIa2y/xWwGAUnzAMz0JN2CveiFnXq07wDYer0kDAP9p83sMhsbeGyoUK3Px5TT72HdByJ1FKxxAo+3NWiIPAAKnz++hjMZmHVOiCMBQln2LKiGXnpFwgDJj74MVTgCAesDMPkaN2OozQorFAJp1O4ZATvUucIAxxpJb/j3oCACQ8Iib5kRTicYAr8rItrZjIcceH5WHSClj8fppfme4xFdWAvzibOs+l+Qaz5fMAaRJg6G+5TyvY8KBLOu5iEDOHbrHCaSRGEU0dlxRTOxc+d5BxV45nuzqPJdA7iV36fZihhWySGsCEWnbi4NVMRi/Mtb96Yn9g6odUptVu08mkItDy2ztfXK+YYWYWLQ2gYgsWX9xqSICELi/nVJm7HzqmIGlQZ/P6w9WHPtKQwKzeNNkArn5pB4bv15dNXQxFUD9zSt+YYiIrG/+wV4RAOIZvrCXomuuroNcrb5MhdadHCZk7DoRpB+efHUHcmnvO5PCqpVZKTvijQ6DuQCbVw25u7xegC07qiwEQCY2iCDbekjZA40Gcmm8/t/LPKqVWQmW1Q4/4+mV0XgsamRL6s0iyOXHd1sYay4bfP50HwAcsE0Esfu0qvBxn3ebzDS+741DqsMa4VkSxTvsdz9lSdctXsjpypMpjrHtyX97Oxm/3gMA4zcJYeLZgYpSdutmg8NlRnTzIxMri4IBn8/r9QZCJRUHtmJ2sg1HKpDjwwsYItLWma8uTSHuulAFgCErhJCtHa0CKGPf6WYWooxhdhsf9YfcP3gXotG1aOYuioi47XQVAPovYyKIXbcFAQD85yxN2Mr2ztt8kA+e35Ha8Ut9F0PuhrM0ACiekRJCuu44DcyBk75rN3jpdDrb6Pd1BPJC7YkV3+2MofWv01QA0B6ICiEmPh5KgKsMerghydA6nU6ns2T/dX7IF8PT5/eh6K/nagBATmoSQ2x/qB/H7B31UH1vipmyNv5uLeSRgVdSQrjzch8AwPBfmBiypjv6E55ZLTnk0V+jKeYMlWasOk6FvLJ6rxh23BwGACj5MCGGyHbdNYBYcRUtWHPSy/viBmUsYY/pOyTRhiv9kG8e3CyG8WerAADUc/bZQGRtz9d5hKyJovnqRJihxyK7Z5x+ZEKGseOWUshDJzWJIf1qJAEAqJ6ZQsS0ACIm1l9Xq9oz+5sTyXhftKtl56IHp5QFPApAYKW9VP0VJZCfTtothmzt8RoAgHbRbrTPeuZfUq0Re6AoqqZ5NIUQsFRuYTbiy84NQN5at0sMse3aMJgHfBa3hYg0tuPt0ytVYkPq8JRIqunlcRrks6O22kD90+EEAEA9crlhz8wSuz+5ui6sqQpxINRrYXTOPbuYQJ47ZIMNxK2X+sAcuORXKoPLjETXzmVP/ebY0bUVUjz1iEjjLV+eU04gD65ewWxg8p3hCpiDl202JAlH/TKUv8cTu186PEAgTy79ImUDsemOGgLm0EVrdE5aXiwgA0KHhFTIp31/6bODdMNlJSYAz7HfdlF0Mi4n/1YubbWDqC+eVmQCUPrdMLeXyUv6CwKAug3MDmJy4dlFJgBQB9Uze+k0J+UrEKDs/aQtxOTCM4uICQL1aDfNNfkLBdAubbaHmFxz8UAVAIZFbAkWEAAjllJ7iKnGv03xhx+n9uLdkd6kwTDmLSAg+FSfBETWtfLTXrSdGEWIWj72modma4UEkKnrqAREpGi/WwMHicebtwH4bm6TIrOFyFJ8/c5ZmOrw5W8AFe8nMqJZsUMUT7DfQffM3xM1EFEflM8BmbjCyICegZrm0RRFUVWPv+qoR5dv74wbDC3p7/M6AM+09YZjmErGY7FYQjcMg1KGtr9Q8jsA7wVbDKec3hXI9wACF69PZVN0cP4H4D15Vi/LGv3MQgBAGfzk7lSWsGdIQQAA4XPndhnZgOu8hQKAUnnp/C4j87qrCgcAUKound2WZJnVM6agAAASmvjQsvakkRHMiHetfy5YaJiV4Oirv9oZSegGk8WoHuvePmP6QK9CoGDVArXH3PTS3I1N3X2xeCKRSCaTiXgs2tHcsOyrp6+dVOFVoDAmiur1B0JFJeVlJUVBv09TFAL/7///uztWUDgg5CEAAHB/AJ0BKgABAAE+MRaJQ6IhIRMJrYggAwSzt3C6qHEWVq32z/c+fpdX83+OvaB33Rs+wH9L+Uf+G+ff+39WvmAfrr0qfMB+2Xq4/6r9tPdl/VPUA/T3rRvQR/kn9s9Ob9qPha/cz9mPaUzWT+h/iF7kPif7P/XPyN/ePzPfS/2z8kP3Q+Aau4zN/ZX75/ef299nO9X5Qagv45/Ov8b+X3+D4kna/89/zvUI9+PpX+m/uH7g/4z05dS/wf/qfzG+gD+c/zr/Ffmv/iv//9Uf4X/Z+L79o/3XsBfyH+d/7L/Gf53/l/5P6af5X/n/5L/Qf+P/C+1b8x/xn/M/1X7x/6/7B/5H/QP9F/e/9D/1/8h///+390Pr7/bX/x+5B+rn31GIQXEAnOLiATnFxAJuN1mTA/B52VIZoXFpTM2pwrqeh1D5waCQRP7lEolRr3KPD5DM6BpPVOXqV7OG/ZcqiVQZnGAiZwO2H9mCT8NH6Ner+6ZHBFzTTzzscUS1bFIriATbzeBYi4r9yYMXmB2+Wb7XwxwdeqUFc2jOZIOUpnhNYyOREtjkOi9yBUqqooKzzeTH+bZR4EUwfyi4isFxdAX2+En2MC4xIB7EKjnvh3SQjHY1OZESJPvFj4pIV7a6hqL+JHoHQggoK6iAEftOhuVWADpp0VBZS5j6WjtQNhvh7zsR1NoPQU6pID1kULTtfp3PQ1Z5AGF1Wsb83uoDKzW7K8fOp+HsdXnp+1xOuvWVEpcqlHJ7hf6/cdnEUPx4v4V0VV25SyOf0p1KVixl9gXPraiTgMGgPVcEoJIlwjqGyiBlpKTxhwXtlbm3ZkjKYv3Pm1VXvmi/lv+pUfEwYZocYsGsWN9fEwcLJSxC1gYZoQt4RDsEGEVLI+Wa5E6TUmyLvPXwVctfFNp6Ez3j5xpXtVE8K83sXDKmlubSP9Lwc6nvBqDxq9BaEoCqeH0OVhrcsbLa0idyzDZQxWBHFQ+5htITH0sjUMDfP/PnalMpoc1kCN2k8zJdZ1AQ/d92gJKB1rY5e7tZ0mODj96XmLjHYiv8zsEBi1Z1w3mJsGM8l53gfvG1TGpuibti3kQuAQBNpqz+QrJR22RBsGSumi3Yf+zznylLCvNphMimd97sAtMrd1+NQP8J4JufLj8XNL1dCR4RsM+Ic+241cNn/rzEkeyBS+4MwZdQfKDxmL+laoXc4wb7o+yhsaPmDHo0A1okqRHitw4//KrZb64hYpv5LKvH3Lsf/K26yn4vBY6SuF+Glp80XTiGaFq8kvZkh77TmonFmb4q3umeu/NDi/6dxd96m1gy7q/aXZfoM0eHyGUWh+Trxt+aXO99c8gJCP/WGBd3Ox0LiATnFxAJzi4gE5kAAP7/6MDgABPpFicp+fNcTEbxb4CRDguT06ppHnWsjjqqhSJNUKC1PTpoRELFoKaPuGMSIn6/96yrLMQjyBrU5EYUEVdN9rm++VBeoACQhq1EPPFxm3vQe7Up5ywdSlKhbgsPQmR53Pkr2X/Pl011FZNp3wP/711cCNMJIJ112Hb1/I5MUuQRM7NVM0a9OeVB70K+WpWYc6gg8rr8jYFhzCyF6QD784yQWSKAlh+RhEms/+A4EDQf0yaNWLz6gn5XNWdpMCh0rrNuELgVFCNS2tn/+w50LmCvgox0/RnE+pi1AulVAenTLgSD5pKq1VNm89FW3pv2LH1PtbRGBvkALrNUuCYci54j1aDo0RMB7fy4MLnhs1+rIzLR5CKyALcPNAi1f37jvpI9YsHt9S2R//qNJE/HtHZ7Gvy3AhFq0jzxLteJXZrSxK0Pm566ZSpFCK8ABpw+qjt2LJ0raOilEqvz0AzOhF2Pts12Jh5J1dhZmYRSgzDJqbDFHPMs33/g7lDdgcbjFlwn+uTp3CcybOkeP+2dhqzKqoyZ5EsLQgVo5b2qjpHssAjjycSHHhXfs//nZgjN5vpBXZl9oSUfeoazQAaYRS9OBWXFEWLIMDEmkWinG5HIxFH+5TW1azJZziTFsOFCPKSgro+HExzUFBnr3hBsclWI8KwrHZq5ddROc5uXxv54iQTgpm1ospjcy9DmmZ/DLRaQFZi+6/hjplMZVe6/H7rN9tw/RsbSnN2Sf5DJ1hx/sQw5/LWLqTlLRFIFdX8sxe5fCVa+HbMzyzHK4jvq2S/b39DSBOTK0HSnP+PFgx7AVcEWRGhOK5jRovESawG1VFr9GGvF34kJBlpPbgUtOxSxFrMQTiKMd+xz3c5SiX3D/X8FAW0B92QULBDFYR/pp8axKcQ0pnKGQ382fs+dj5XJ3J9v1XfYG3IjvcNmEkxMkJTEwTLPd8oAaipR3w3B/s1gbxyQC3Oz6qN6b8zkqg9m7bTAzIyNhHxkJsCuxuVbk7GrtqbH0+wSnFMFXrPY0fq+Jsq9TGCIemAVXCXv75bWsZ9BgOkS31o3+5tuoFPpz5d6DPErMFjkmsRiI7dLvphBE28xfbkdnV4wHxL7+1sonxguMs/CvusVXw/stHxj4Iv8N/g4Jl+cefFIkFOHkiaSwL4gdE4Gq+mDMp4lPeR9DJddJo3xeVpOfP2Wr213B/g09M4Qgq/jQ+obbr5iyZ95GX0N0vR5cBJF1sAVWuvGrk5jufC2TXK+WREL5PIeUwovkgxkOsgirBiI71NM+4PavIe01mgZ+UUiRAjHuVBXZSH3GcVBGrA96twEcXvKIh4Kis4rLiGRQHmMKJUEtu/zWG23R/JP540DMCzRcNdNJPON6G+cQXfTfAJuwabHD7/4Lhifq1NsP5ub0M4n/3cRhcUYegkIKygi29i8xvr51Ff8vwGK7+LXr4kjZrKp8JGUOPWsNP11BBpZwb12NTizYDy9N39NcUXagEJwSd2RNkAQ84GrLAfkRCECIRo6v9jmGnM1eh/VASZxLU0385F3tKiJQ73uRJw/v+c5YjhhQq1jXT33chKhpqpKVXRBIlaS9PTdF9rrXHYGSrpP7noCxu2eUruX7uQ8QjSgdn3Po5dJ1s3qsPUlUW4v2nJNGteF28Ck5ZARDDYQUXJ1b6GYCKAAwEpyVbTDI2c5OPEe8zajk3Gjx9wpUIbqOHoczTj3YEJ71iuK8WsrolyuQi7f0OcVNtCH3IM6C5jqkWvFjBH1RUXz85xSwtsNkcMNYYpRThZbMzhkD8+7QNoAaQt4eXBthlbkHDei/8cj+5wdLH8MBBJgG1k0RQSkkTLYmgdJiyQUwpqfb34Tqh79oYRh43Tdz/mw0EJtj+Ii3Y9jPp72+SyOzZ8JV67S7J9TKbOUeHxP8I//LV1h6P5YtHlYCb9ohfNttdagzDggrn3s1hHEsH806ii5IJVfBvIOS9DN0/hIVSfpQDtBMPgao9PTnLD9+9lqj/T41DKyUvvKaa4Jhlnjqa1HwsXbsUIJaRRoWsBuCUgEukEE6Eo/AYIwjv0jqTiZhY0+QXssEq2Lw8PA8/NV7p9rUsD07YGRfTHesmqaGaiZj+2KhEBQnws4nnTG9UHeELDIDqfLS1eekiXrmQ3G3ZgQzGyfO9FHTbuw0E8+UjNKudq4AzlEXxLx+CiqjOZfi3ttHl+oOcyLVHr/OGgeC9E/bs+bWhxadgk8BFh+ulH1hb0gD9AbFJC8YO5gSXYf/W3Wbj93YgpZv7cFj6ULv/DFnShznnEi0napCR226OrOas3fCQUxaA4vyusdeNy/532qcazq/POp/2lm4QbDDAb66JSiw9SWYxHkL9PMMsblZi0oPclzK7s3h3uxx9OEjrdF/y1Q5PF5HltnPah6hvxsLMVul9AZbMtYs193loqpx04WTqjoJGHJPDa/M1eKMfx1/z9HDdwkyMMMjLX1/ZxxXXuJ5HYQ/TZB1ezIA0FOKoBqkFqJqI7J8yMSPurh4p5UqUrMnhInlnqhBEITAnsEZiyVJgu8aKzjT8oWiBeYGD5WXNm7XSdu/OisaEqfvvV/IkloAO/mLAa+GWYQ5pZ94vhBY21eqb/obqlky3GT1BgDaf/57gYj1azfUPI6gPl8/mCmowbXefbmRl2TmwRMA36LSoHk6L/6SDk6oe6mntgNdVszO2oiKPYMVSO1UEl5b5ExeYhHQsnVLcFhkSUnW8Rvtcxe4dTjEaYlqg1hY82gSE3xe7s1DQCWeYDr8fAT+CGYROxlsSBBVekb/8W0r7qZC+PMZxZmscFK6tnbVF60oPfkTp3TYv82WSqO5O4cikqWG8zrSlKUjnaUHnRYUWU95O2gpwDorXnyWtFzP8K60l6Acpy4Q/93KhoGJuDcUfIcAxsssorfu+ObhS4gwEn8OfLwIc6904WdvPyR37/4nJHF07JSuldIs47mEoL18Fow3yJ/WDBhi7scZCBvzqgiuMJdKlzz2znJyIljETdWT5GlqTJkK8Rdletk1GitWoiwhnw/tx/cXx2ZfZb6gasVkczvDEhjazx4Hzuf+G4ET+6Bg7QuaMC3JSAZksmnIF8gwK9V6H3uFefsG1Q8Ua1Yrz9PsWeYrqk56RT/K4P7Jnw3T1qsclcg9NFkw5GT5c+zH1YPKjFc6bOYeWh6e4P4Wvp/PEsF5bgOUYIf9tS26qEKtb3X908GdZV5hUMzMhpCmiiVTO+n2tna9JxrJ5fJ3a9Aqzbby9uJPWeSvS4IX6i2uD/o9VzNOUkiuxBiMfJp3MV4dWYVqUn1UDZSesyioq7Auqq5mYZ6FE6S/Vm2sGnguEPVz9mbyjLVculVrHvUpu+7524+vw96GoLwnQG9ML9HvMYuNUd5pTxqebJzJxPjZTAxQQL1oUZh26qFoUMd42YL8zoElDOxTYrfXr2erh4AYw4Crc8cA35pi+l7o+3m1OMWAh1i7at1BDZn0kbTGIChZuk800X1iYxJ5R4FFgb6ezwuaNkRxh/QWn5Rmx/BVcrX50HuLYgN3I8nqmzoX6NlO4s4RosSiKgJ2zXghIELliHqxZ2GsHAzwixr02Pis0J661MnqoYuKxUWM7OG339D/onrugk9cV+JaD53P+jiBmqrb3Wu0KrIQMhKXiV3rwUwbL0jKF0iCUTJqVYOzCB2JNUpNfZUlSwP85QfW82dVcgpIOrJvyrnxi7GXw64MddCaD3kRyxWXrGLnDu3ej2Kba1vmZs6so3lzeTLdG2oUuf4JbLG0pxT2Eh9Cs8Jg9wXBRAMz43M10heUXG8gAeSvkgb+mnAQwArMtGOuwU5Tv65uhMfQqV3T/oOMOfsCy7rcqo6LAjWbKyUQrhFv2KjT+zTJ+KcX7S/ES2S991UuPLyX6mTAUUKir1PFW8xZlGSrWFhNemjQnvHoY1Eujr+Ic9pcE87egtqy4kBtmLqq3oBdfj/gW3UocPZ136MefCyV35TQaROLaqCdSK4Qtzk2BC7SMWZKhESZg5n8LYYpg8GKYjjRaD3vyqPpAE8IcHMBxkJA15EcFGgT5qKiOMHeyaz5h7AxzCJam6zIe/cX1zejvkhqs3ONWbEoiqGJQQ8OB2F/ASCoPQK4JgQHXhUvPM19qvTML4iHaVEUvzognrnUZVWPy8Xqz9cSBIttgUUoWZZJ7T+8vTXD9c0ztc7tsuT2ICoXUiwt0XNQZX1b0tqF6i9PaX+8x9ErYHCl+10UrjaqIgU9PRC2UT+45hWd3I5rAuzQjvz9r3pZg+N7b6lrfqYTtiVOxSkmX2Sm4KDtcJgCY3ULJddvrzgR7R4YTYjSc3SM+42gAHzuXb+6HFk3Su9WmBdTgarMo3GR2PSyfL3vAjDKXYR3IWaEmVDakWa+iXm/xrlhyXgBRWCWq85+9oqxWlL5ZYdcIiS0aXqCMtQCEUJthwN+hPKFkR8/1YTH2p/kpc1z7cxxd/pAPof+MX3r8/1ukty2bv99z+KnpUiQmHsxLlex5ZxuGUW3Tnf3MweUIsWeKe4VlQy9+4SoJLyhFs9bhyt61a2kw+YC0Rvy7YEHcMj3e2sR8gsy/fYDSqQ/TU4g2A08tPNSJt0Bof+hwY/7/PUfFo2xp5wHuN+zIVpMohH5y6crf6IwXXGIfvF8O/BBRYRfCKwnX6cKtYG+iVmqRfSRnw52C5LbgAF49SjKZBEKA+k86zzlSsu7sqK4ZbvNm19nHHCFlRzUI98D+ksXJfLY3yJUyzjUFZ1DrlZJeChgb4s/MXW2qio278D2ci30jb8bFxxAlEK4XQV17G+2JVF6igfGKRBh8iquj3wnA3ztHGd3IafJRZAaPJ/pM5Mk2PU+JB6CPqaCFzKKnlPvW10bLKtVkfG+v/PEYixJPouwyVMpULmMAX5G34JjgbWU7n86C/+UlYKfQdmru+RN3HyR5MIBpoW8rsVmzOtwQMdkJavWU2XJLElUFXndzMyzyVBm++O0ECgetTgXB9U5qVXi0ILTAR4yNcER6cboB4vw/iypj9E4Mb+1WMZ7ewYhOt4oYK60Jpd22ejUo9ppCYxr+h7r/aYPTKWT0yAExpMQCZB54mX9U+DhCRO77lkXEjzFx14vZFipuixnJsKi8xCTTr9noQ9t+lrZeZ72yihciktmdj4GeymYtPfI3KEwUyjkTyzhVm7mVbDDms21HqN5Z5L94TS5w5cubdaW/B/HuWbbsA9KdwQKuOmOzTnmbDhvLiUrwp4S8rGllcy46t/NBqrQAuIDWR+cahpFcnLXgR0wJgSjOOe7wF1YKz1EfSTLE6vOOe6SPWm5wehB8O2zJQ6X2HNc82e7bmlCNr+Pj/xBJDWqrN5ir0p+s7FJ3DxUaUQMZebG6I7p/7z1erA1ThDtPgp2cdb6Xu4Xy27J5nlqD4tmvB3UEclEGG0dyRmpLGQD+2SPI/YNZ6z7wMjM8ucU941rF7SmTDK+3bPKzzOIAKmsz61hpoLhIfJ763tcfX+EEMlnRIBnzt9iS8PHBsP/gb1g21hweBBLy8Q1A6uXR49cg5GVKA+D0hXT1XPPFcy+h+lPffE/IkPYwqCbB3pkNGamVhof5CZL8w0P/uv43I3+uspAnfnT9tom9UkCGHGDdId+drG+NpNmd120mxnVeOvxh45sLyo2o0yRLfLZ+GQ68KQH32FFG8XOVOFFSHYfA0fHfEH21q0Z0EIBYH6PZCm8upeAPJ2bSE2RuSvYPtTjlnfb77QMTDvv8ENGVhsWEblEfObpNgbr3VYyhIi/ZX4SPn/u8PGM6oHpK0Wnup88sW6j7+vbO7gek/e3Y/kktZkU5i3WMM2VV4JDAErz9w6rg9dkSov0GG9RFRVSgD07wtvtUKDIGBsxuDhYWK/FaTilkD/1WPzSOBAOjGajIOO/6uLoEAlDItQuTTRGgLFkjUhOlhEfUx1sWLruRturA7HwiVWAdUZZhiEs4CIEZ+6f24NoHjNYYZu94+sfwoSRheu5/CPYjRfr3iPL4iAFVZK2F5j/k/Co0XB7b37ynpXXldlR1ZTY8X4DMfwoR+hpJhuU34Pju0Dd5uWMKHqbxhOJYpytsirdoexL/0okVn7QUtwsE4QuXPi7sbcP2tJo7pymhk6bCEDUlH1HAmjLoeHIHdTPX1x/O3jafS4MczDclemearJ0j3ZhdBE8pB4Ex4ypiKYbDn71Ow9V3lUFfP8Kwd1ZXOCV8c9IGC8K1ADJl9q9IHdlHvuaYfGDfw1pwIlfe3CE7BcEEaGdQY32bBgSWuRolJNBMmCcQHmyxkt1IP1fJkzTdDjZDpmshL7wYIx6QZsrxO1KXvwtvruIXLQY0UV/IxPNELroVd9ZiZfeFFgsVOgRuGsXeQ8bo5LLkifSLvSymSj54EXnm0GfzLhd5EXNpo6w1Fwuu5Yv3fmZZfx/DJ9W/DijEiPYYuh+TPVI36dVOAfLEVaWDUF62YZlyc/uVw8gPYPqsx1SfDIXqg5YS9XyTXypnSBCpTUunojJpa17hIQh4ivC+xG4p3f/pmz8N2Z9OkIxdcKV7Xj8ZqUD+tg0eN2pOW+Ctcx1TKHEQtTOPcKsyO/q6VPnlD+ziIAXbTo/61M+rEWgxnEO2e7ZRzxlXzDV97m0MX4JcGmBkf0aAm8xUyKH+0I85lW2bIQBHHIzO8A7ABzfftkfABDUTh3a740+1xD2iTIY8PJNDFyT1ZMYdbGGAchA9jn8Arlzlzg3B/ib8Z+NBiYR2mbO47PSr8yMDFPYY00ORY+D1UH4ZgnrQ/redwzRQ5TuYGo66q5ULGIxWUwhJIT0rbDU4vspuD71EkCOCCgtY9usnnfQz8T6Mo+E5FMTxFr6eSLGAfU060oGgFnsxDVW8+yDcv865Hks9CpnePP9iS2EmNoCf0CGILEXA+CkkuKZ7WBr+m8j7NoiEUBzzwKeQvQHGNYTrP35a63jbT/JkrGgwGDI+liAkdptZ4fXia6VshhBeDxDr6uyHgTgWefPud1/o8pbufROVj8ZqQ0V/d83sqIdKmSXTeHyCkW0yimLx5sZNNpQi/CCVqjzIGyBe+JXbgX3hVvIw4SBy8sm0w+woubcBBgIvTXtDw88e3ToC3LBAsbwXya5Iih65YtrKW/ElHDgZFxYcfnvfOGK52qJdRDX/DzDGeKKGHmCVnFdSJEYOFo5WwO8al9zJI3FQuZxLB0pDzBVVEYyYnjuNPcDv1uXndIsd58fbBbhChN65Im/bN/I8O9BYzLRusOOV1XksQZ9Q0l1uj5w8pDBr9Ek91QHJBZKj+kN1tD+xk9umqnRYv123BsKUwxvM7pMaf1X5fmNT5BcMvDHNs2WLLO/0fooxZbpimy2+iyt0/UZ2cSs9vx3BPWLgvCD6frPfU+Unte86gBG9T6Ue7r9eJtnQ1TQ6dCzgH23GTW2SieCDZYVw7ftc/18LIBHCiBMgCQ+TiOu0xMiBeQ+B3vqPIr5baRFR3ecASgFlPmeocE+leJJjeqEJmMMmdpHHpt0+wktKdNI79Np38c/NgPSdFWV24fBRKiKOsS6oc6JXoIWzMl4KEOI2Tt18fCgKZ92AAWfy1KKuZIpQJAq2KwZ/KTDyII5atMXIEVwl3VBVQ3ObB7HTah2ADb3c5ZqEu7a7Ij7ih+LmLES6BHpvhEoNxBcE6EWIItgK6xUPmeOO1EOk4Cd6B5zZJhFjMCQVwVy3vEE+pqCOccnZ8glXqR9JUfGc08ifW+s6nx6+zlKb7YpOUVexPDU0SUpqyDE8XRuONiBx6S+IOhw2mD6w8g+eE4+URnLRyiFI0pMzFNdptRGsyCDbhsFnyZezJCIwHSADypp2QVPPof67xDXJz4LLenAjwKm+B7qcCEreENA4hfh0U/x7XC5blRLyRqaK9qU3RBkvmyvI2+8GCVbrxV0SlP7cGcsQ6jR8i6tUqA/nodkB1LzT0ym067O0X14Xr/auaGK5DbzdT4yAlI6YAzYHJuCJkhDlkzGW3yw/LoGtE3/7uZzY5zThT3f2Q8Bn/+GIP2kliV+wNNgu1FjJH3+7gg/oMxqUtqtyve9JC0Nwtk3a8AkyUnL5lyN3ii0mK49xWJNzQqYt3bhxn129MXGCMDgMayT+TscHEQfXNn6PbFhM9xI3mTWm1CGD4nNTz6ZzvE31NlUEefsE9N35q1wqWDKMv6zcf17uaiStFaCdCeXfiZn9mImmKFsamVwzY2l8i3wINIvYbfif3XXfTOZeJE/FqG9H4bcOTBRjpHphZAYzUdHm5aLJni1uHRk4IqwTUqoLuNXDgBwhJOyuPJc7w4Ay18d71q1+Y3EQCXh1qqz8O3gK1brpLEV8qXPkuT1c5T+mFvVLcORvGZWz2pgzHuOgXD5ICz5mcWE8gp/ex9wEyblZl0kWoQIeW1pQURxKPdM+ghfxGRF5FpLbAyHinZ+Q71PbrXgT0tCjh+bp7sb/d3mCxFYhGjgY4iiykyGWrXWjE7BQGplhB+T1FD/tDIe3Wa3bfzrSJuNkSXDb8LYWPHYWEmKqPwLXFJIB9g8nJx2gWeD6/yB38YwDrkXXfo1K/LNs5kHO+TXpv21jNh9FVOVg2dn+NUfrO1e/wRcCfiRW6TYGlLr3bKX+vVGvTLbqM+dAzwUhRGHmmv7249RPIynJ3i2r2OAVrS/fsGv8+Iz1GXAy3PFz72TWLx1sb2mrpUuQwTtF2M0rI5gqfOp1kr3naDHuVbo4tHJo3R+zr0Ll2UxBz3/A5pwVdsxrztOUNJ3cJDFEh2n7maZfigg9aMvSRzv+b0gnxA8rZf+oLe1X+PddchxqEASdIkGdS5mcGgL1PUBiQ5YCX1N6hYDg2BDsnRGCHi5xH6LnoQexQVTv/iZb9RUfmEH/NtPS9VKN5m4d3py+RDGgfuO0D8GRxjpwOyrf9/o13y/ypitO3/7vrvJuWtQ1QVuYvyIibLIkMH81oLTfk3pSgKr0JKqlPPBqR8JpdPZ4t5EQibD8NSrkCBEIRSybN8220MYkFIGV8vQhZTQ99Y1o333qC9XMyAl6wgxR/3SXK8ihzAAmrMn2sx7N0h6LLCf7RNf4oqrZ2DCrXSZyi1cU5BOXXe9myfCGSAaALSW++BEiPPnZY9pCFmVWnvpNb5SoAgwWAQ8VfUY1RLgCXxPB3XVGMnl8MwpNpEZLcGnBu4z2YJXJBrIiTG4lQU+YWUwNyJJXIhaRBysrCfO9NEcnVUkzitHJuPeZZsfiMGKEaOon+GC7Gbj93Acygy6dBPFwZ4bloRiDmdVdDeL1oj83uLSp6I3qEYCflWNS1Mlb9TEEwOqZgHU/wdTaFz0MW4bO9KShGSyAnl6p08dZIXTBN3+AkFacS/tHN89jjvPEDz3SrNMVVEbT9aewMAhMByivil7pYaTxq0Z44J5el1Wy0AImSCzpnA0nTmba9Yyb8UKagQyP6yQvAh1N8s14ElRvL3iLhgiMwKxc6eaxKryKFPwSqQCKdQf856Nm3tvoWggH7FPIdmIOdHoSnucDsVbFo8g22Klc5S9cmdqIiRrZkWu23kxZURHmiK0EmTG054f7EQurdrQ0+r6FIJMM/55jbWWoG4ZbJ4An73cfYItVYyl+ySHHEDyu4z/gF+R+jsaF5Xb3lu1jI4kBkKcT0/O66eDFQHIvTqjcFFNRO9Pe/E/oN1Rii1S8fVJ5BeHP3cKup3ThMktvOzdMjFrfntgpCTc3JGlzfjLPOLcDo/iTsv2jyQBl4BYZrZxnmr9r8F7IzF97OH70eqT4QUb+owzEKURfFazfTw6tctT3aoe/U1tD+M4NPLbsB8tgLV3FNh3D+y4QfAiQwHNywItFxCtV/FGKKP6n+TQOdn+DeGrBrFmnUwOY6Vc39PskI+ek1cGlj+0oXTvOyh3aoeeo+RSxiRiByyunyNOkpYgEyFWsGz6rR+PMWeJv6ldfRKawJm9imufpeJQUKlsMgeSiI1DDeweQTF/+1DRtTbgykt5PRSY+IuEaDMC0Pri7QEY25PHOfxhnwjak4FL0UuQAAGHM3S7JpTBRAIpJHoII7SFfm53XbB2/Zn8O5X9w4wNaifEcgWSowCnDQR7pzIVnhBCxxMA9eL//NPmqoMeX7G3EVU1yMYCS+yzdKEDFxWbB5mRrc7Yv/TgzyQYPaSXczOx3CUxwS54ADFIYcI/ABJm7u9ZeQECrH7nmBbyGR3dRMXEufZEyAAAAAAAA==", "_w": 221, "_h": 221, "ue_properties": { "widget_ue_connectable": {}, "version": "7.8", "input_ue_unconnectable": {} } }, "widgets_values": [ "{\"filename\":\"Logo no background_00001_.png\",\"subfolder\":\"\"}" ], "widgets_values_named": { "_image_path": "{\"filename\":\"Logo no background_00001_.png\",\"subfolder\":\"\"}" }, "color": "#fff0", "bgcolor": "#fff0" }, { "id": 71, "type": "LCPersonMask", "pos": [ 2450, -3030 ], "size": [ 300, 490 ], "flags": {}, "order": 28, "mode": 0, "inputs": [ { "name": "image", "type": "IMAGE", "link": 99 } ], "outputs": [ { "name": "cutout", "type": "IMAGE", "links": null }, { "name": "mask", "type": "MASK", "links": [ 100, 101 ] }, { "name": "raw_mask", "type": "MASK", "links": null } ], "properties": { "cnr_id": "ComfyUI_LC_MaskMaker_nodes", "ver": "a8c3e1c6a19abfae17932561c88b32d41cf426a8", "Node name for S&R": "LCPersonMask", "ue_properties": { "widget_ue_connectable": {}, "input_ue_unconnectable": {}, "version": "7.8" } }, "widgets_values": [ "segformer", "models/segformer_b2_clothes (local)", true, false, false, false, false, false, 0.4, "vitmatte", "models/vitmatte (local)", 10, 10, 0.01, 0.99, 2, false ], "widgets_values_named": { "engine": "segformer", "segformer_model": "models/segformer_b2_clothes (local)", "face": true, "hair": false, "body": false, "clothes": false, "accessories": false, "background": false, "confidence": 0.4, "refine": "vitmatte", "vitmatte_model": "models/vitmatte (local)", "edge_erode": 10, "edge_dilate": 10, "black_point": 0.01, "white_point": 0.99, "max_megapixels": 2, "remove_features": false }, "color": "#324b4b", "bgcolor": "#324b4b" }, { "id": 72, "type": "LCPreviewMask", "pos": [ 2760, -2740 ], "size": [ 180, 200 ], "flags": {}, "order": 33, "mode": 0, "inputs": [ { "name": "mask", "shape": 7, "type": "MASK", "link": 101 } ], "outputs": [ { "name": "mask", "type": "MASK", "links": null } ], "properties": { "cnr_id": "ComfyUI_LC123_nodes", "ver": "e9d3f5fa5d7d6c0f603b2e19f8a8bf458ee3b8ec", "Node name for S&R": "LCPreviewMask", "ue_properties": { "widget_ue_connectable": {}, "input_ue_unconnectable": {}, "version": "7.8" } }, "widgets_values": [ "" ], "widgets_values_named": { "lc_preview": "" }, "color": "#222", "bgcolor": "#000" }, { "id": 73, "type": "LCNote", "pos": [ 1190, -2440 ], "size": [ 430, 890 ], "flags": { "collapsed": false }, "order": 21, "mode": 0, "inputs": [], "outputs": [], "title": "Relay Notes", "properties": { "lc_note": { "source": "en", "texts": { "en": { "text": "## LC Relay VSR Upscale 🧩\n\n**What is it?**\n- A one-step upscaler for video and pictures (2x or 4x). The model, VAE and sampler are built in. Fast, sharp, and it adds real skin and hair detail.\n- A fraction of the time VOSR2 or SeedVR2 takes, and higher quality.\n- **detect_faces** finds the faces on its own and holds back Relay's grain on them, so skin keeps its look.\n- 💡 If you used detailers or want all skin protected, wire a mask into **protect_mask**. It overrides detect_faces.\n\n- ✋ **First use downloads the models:** 5.2 GB per upscale size (10 GB for bf16) plus a 128 MB VAE, and 55 MB for the face finder. A toast tells you it started. Press Stop to cancel. Already have them? They are found wherever they are.\n\n---\n\n### Videos 🎥\n\n- **2x faces on** is the default. Your eye goes straight to the faces when things are moving, and everything else is background noise.\n - Use **faces off** only when time or file size is the problem.\n- **4x** runs on an 8 GB card, but on a moving video it looks barely different from 2x. Extra sharpness can actually make it look worse.\n- Save with **LC Save Video**'s **relay** input. It saves a few frames at a time, so the upscaled video never sits in RAM (5 s at 4x: 0.3 GB instead of 12 GB).\n - Leave the Relay node's **image** output empty, or the full video gets built anyway.\n\n| Video | 2x faces on | 2x faces off | 4x faces on | 4x faces off |\n|---|---|---|---|---|\n| 960 x 544, 5 s, 24 fps | 13 s (est.) | 12 s | 36 s (est.) | 34 s |\n| 896 x 1152, 6.6 s, 24 fps | 26 s (est.) | 22 - 24 s | 78 s (est.) | 72 - 75 s |\n| 832 x 1056, 15 s, 48 fps | 46 s | 40 s | 2.7 min | 2.6 min |\n| **8 GB laptop:** 960 x 544, 5 s, 24 fps | 28 s (est.) | 27 s | 2 min (est.) | 118 s |\n| **8 GB laptop:** 832 x 1056, 15 s, 48 fps | 4 min | 3.9 min | 18 min (est.) | 17 min (est.) |\n\n\\* RTX 5090 unless noted, laptop is an RTX 5060 8 GB. Models already downloaded. Faces on finds the faces every 3rd frame and costs about 4 - 15 % (measured on the 15 s clip). (est.) rows are scaled from the measured ones.\n- 💡 VRAM stays flat on 8 GB cards (7.4 GB at 2x), so long clips do not slow to a crawl. The 5090 used 13 GB at 2x and 24 GB at 4x only because it had it free.\n- ⚠️ **16 GB RAM:** the frames sit in RAM before Relay even starts. The 15 s, 48 fps clip peaked at 18.5 GB.\n - Keep clips short, or upscale **before** frame interpolation (half the frames, half the RAM). Recommended, not tested.\n\n---\n\n### Pictures 🖼️\n\n- People actually zoom in on a picture, so 4x is worth it here. Faces cost nothing on a picture (about 4 s either way at 4x, 6 s for a 1056 x 1320 picture).\n- **Quality order:**\n - 4x faces on\n - 2x faces on\n - 4x faces off\n - 2x faces off\n- **Realism and photos:** faces on.\n- **Lineart, anime, painted, etc.:** depends on the face.\n - Smooth or painted skin: faces on (the grain is just noise there).\n - Freckles, hatching or other drawn detail: faces off, or that detail goes soft.\n- Small faces in a crowd: on or off makes no visible difference.\n\n**💡 Tips & tricks**\n- Adjust how much the faces are protected with **mask_protect_strength**.\n- Adjust the sharpness with **detail**.\n\n---\n\n### Settings ⚙️\n\n- **source:**\n - **video:** the frames are matched to each other, so the video stays steady. Use it for anything that came from a video.\n - **image:** every picture is upscaled on its own, so nothing carries over between them. Use it for a batch of different pictures. One picture comes out the same either way.\n- **upscale:** 2x for video, 4x for pictures. Each one is its own model.\n - **2x bf16 / 4x bf16 (10 GB):** the original release. It's an option, but not noticeably different, and it uses about 5 GB more VRAM.\n - Quality: practically identical. The average difference is 0.12 - 0.21 levels out of 255, and the 100% crops can't be told apart.\n - 4x on a 5090: pictures 4 s either way (10 - 11 GB vs 15 - 17 GB VRAM), a 3 s clip 22 s vs 25 s (19 GB vs 24 GB).\n- **detail:** 1.0 is the full model output. Lower moves toward a plain upscale.\n- **detect_faces:** typically on. Off for drawn faces with freckles or hatching.\n- **mask_protect_strength:** how much of Relay's detail still lands on the face. 0.5 is the default. Raise it to keep more texture on stylized faces.\n\n---\n\n### Manual download 📂 (optional)\n\n- The node downloads these on first use. Grab them yourself if you'd rather not wait mid-run.\n\n- [Relay 2x model](https://huggingface.co/Comfy-Org/RelayVSR/resolve/main/diffusion_models/relayvsr_x2_lightx2v_int8_convrot.safetensors?download=true)\n\n- [Relay 2x VAE](https://huggingface.co/Comfy-Org/RelayVSR/resolve/main/vae/relayvsr_flash_vae_x2_fp32.safetensors?download=true)\n\n- [Relay 4x model](https://huggingface.co/Comfy-Org/RelayVSR/resolve/main/diffusion_models/relayvsr_x4_int8_convrot.safetensors?download=true)\n\n- [Relay 4x VAE](https://huggingface.co/Comfy-Org/RelayVSR/resolve/main/vae/relayvsr_flash_vae_x4_fp32.safetensors?download=true)\n\n- [Relay 4x bf16 model (optional, 10 GB)](https://huggingface.co/kopper/RelayVSR/resolve/main/wan.safetensors?download=true)\n\n- [Relay 2x bf16 model (optional, 10 GB)](https://huggingface.co/lightx2v/RelayVSR_X2/resolve/main/wan_x2.safetensors?download=true)\n\n```\nComfyUI/\n└── models/\n ├── diffusion_models/\n │ ├── relayvsr_x2_lightx2v_int8_convrot (5.2 GB)\n │ ├── relayvsr_x4_int8_convrot (5.2 GB)\n │ ├── wan_x2 (2x bf16, optional, 10 GB)\n │ └── wan (4x bf16, optional, 10 GB)\n ├── vae/\n │ ├── relayvsr_flash_vae_x2_fp32\n │ └── relayvsr_flash_vae_x4_fp32\n └── facexlib/\n ├── detection_mobilenet0.25_Final.pth\n └── parsing_bisenet.pth\n```" }, "zh": { "text": "## LC Relay VSR Upscale 🧩\n\n**这是什么？**\n- 一步完成的视频和图片放大器（2x 或 4x）。模型、VAE 和采样器都已内置。速度快、清晰，还能添加真实的皮肤和头发细节。\n- 耗时只有 VOSR2 或 SeedVR2 的一小部分，质量更高。\n- **detect_faces** 会自动找到人脸，并压住 Relay 在脸上的颗粒，让皮肤保持原样。\n- 💡 如果你用了检测修复（detailer），或想保护全部皮肤，把遮罩接到 **protect_mask**。它会覆盖 detect_faces。\n\n- ✋ **首次使用会下载模型：** 每种放大倍数 5.2 GB（bf16 为 10 GB），加一个 128 MB 的 VAE，人脸检测另需 55 MB。开始时会弹出提示。按 Stop 可取消。已经有了？无论放在哪里都能找到。\n\n---\n\n### 视频 🎥\n\n- **2x 开启人脸** 是默认设置。画面在动时，眼睛会直接盯住人脸，其余部分都成了背景噪声。\n - 只有在时间或文件大小成为问题时，才使用 **关闭人脸**。\n- **4x** 能在 8 GB 显卡上运行，但在运动的视频里和 2x 几乎看不出区别。多出来的锐度反而可能让画面更难看。\n- 用 **LC Save Video** 的 **relay** 输入保存。它一次只保存几帧，放大后的视频永远不会整个放进内存（5 秒 4x：0.3 GB，而不是 12 GB）。\n - Relay 节点的 **image** 输出保持空着，否则还是会生成完整视频。\n\n| 视频 | 2x 开启人脸 | 2x 关闭人脸 | 4x 开启人脸 | 4x 关闭人脸 |\n|---|---|---|---|---|\n| 960 x 544，5 秒，24 fps | 13 秒（估计） | 12 秒 | 36 秒（估计） | 34 秒 |\n| 896 x 1152，6.6 秒，24 fps | 26 秒（估计） | 22 - 24 秒 | 78 秒（估计） | 72 - 75 秒 |\n| 832 x 1056，15 秒，48 fps | 46 秒 | 40 秒 | 2.7 分钟 | 2.6 分钟 |\n| **8 GB 笔记本：** 960 x 544，5 秒，24 fps | 28 秒（估计） | 27 秒 | 2 分钟（估计） | 118 秒 |\n| **8 GB 笔记本：** 832 x 1056，15 秒，48 fps | 4 分钟 | 3.9 分钟 | 18 分钟（估计） | 17 分钟（估计） |\n\n\\* 除注明外均为 RTX 5090，笔记本为 RTX 5060 8 GB。模型已下载。开启人脸时每 3 帧检测一次人脸，多花约 4 - 15 %（在 15 秒片段上实测）。（估计）一栏由实测数据推算。\n- 💡 8 GB 显卡上显存占用保持平稳（2x 时 7.4 GB），所以长片段不会越跑越慢。5090 在 2x 用了 13 GB、4x 用了 24 GB，只是因为它有空闲显存。\n- ⚠️ **16 GB 内存：** 在 Relay 开始之前，所有帧就已经放在内存里了。15 秒、48 fps 的片段峰值为 18.5 GB。\n - 片段保持简短，或者在插帧 **之前** 放大（帧数减半，内存减半）。推荐做法，未测试。\n\n---\n\n### 图片 🖼️\n\n- 人们真的会放大查看图片，所以这里 4x 值得。图片上开启人脸几乎不花时间（4x 时开关都约 4 秒，1056 x 1320 的图片约 6 秒）。\n- **质量排序：**\n - 4x 开启人脸\n - 2x 开启人脸\n - 4x 关闭人脸\n - 2x 关闭人脸\n- **写实和照片：** 开启人脸。\n- **线稿、动漫、绘画等：** 取决于脸。\n - 光滑或绘画式的皮肤：开启人脸（那里的颗粒只是噪声）。\n - 雀斑、排线或其他画出来的细节：关闭人脸，否则这些细节会变软。\n- 人群里的小脸：开或关看不出区别。\n\n**💡 小技巧**\n- 用 **mask_protect_strength** 调整人脸受保护的程度。\n- 用 **detail** 调整锐度。\n\n---\n\n### 设置 ⚙️\n\n- **source：**\n - **video：** 帧与帧之间相互匹配，视频保持稳定。凡是来自视频的内容都用它。\n - **image：** 每张图片单独放大，彼此之间不会互相影响。用于一批不同的图片。单张图片两种方式结果相同。\n- **upscale：** 视频用 2x，图片用 4x。每种都是独立的模型。\n - **2x bf16 / 4x bf16（10 GB）：** 原始版本。可以选，但看不出明显区别，还要多用约 5 GB 显存。\n - 画质：几乎一样。平均差异只有 0.12 - 0.21（满分 255），100% 裁切图也看不出区别。\n - 5090 上 4x：图片都是 4 秒（显存 10 - 11 GB 对 15 - 17 GB），3 秒视频 22 秒对 25 秒（19 GB 对 24 GB）。\n- **detail：** 1.0 是完整的模型输出。调低会更接近普通放大。\n- **detect_faces：** 通常开启。对带雀斑或排线的手绘脸关闭。\n- **mask_protect_strength：** Relay 的细节还有多少会落到脸上。默认 0.5。调高可以在风格化的脸上保留更多纹理。\n\n---\n\n### 手动下载 📂（可选）\n\n- 节点会在首次使用时下载这些文件。不想在运行中途等待的话，可以自己先下载。\n\n- [Relay 2x 模型](https://huggingface.co/Comfy-Org/RelayVSR/resolve/main/diffusion_models/relayvsr_x2_lightx2v_int8_convrot.safetensors?download=true)\n\n- [Relay 2x VAE](https://huggingface.co/Comfy-Org/RelayVSR/resolve/main/vae/relayvsr_flash_vae_x2_fp32.safetensors?download=true)\n\n- [Relay 4x 模型](https://huggingface.co/Comfy-Org/RelayVSR/resolve/main/diffusion_models/relayvsr_x4_int8_convrot.safetensors?download=true)\n\n- [Relay 4x VAE](https://huggingface.co/Comfy-Org/RelayVSR/resolve/main/vae/relayvsr_flash_vae_x4_fp32.safetensors?download=true)\n\n- [Relay 4x bf16 模型（可选，10 GB）](https://huggingface.co/kopper/RelayVSR/resolve/main/wan.safetensors?download=true)\n\n- [Relay 2x bf16 模型（可选，10 GB）](https://huggingface.co/lightx2v/RelayVSR_X2/resolve/main/wan_x2.safetensors?download=true)\n\n```\nComfyUI/\n└── models/\n ├── diffusion_models/\n │ ├── relayvsr_x2_lightx2v_int8_convrot (5.2 GB)\n │ ├── relayvsr_x4_int8_convrot (5.2 GB)\n │ ├── wan_x2 (2x bf16, 可选, 10 GB)\n │ └── wan (4x bf16, 可选, 10 GB)\n ├── vae/\n │ ├── relayvsr_flash_vae_x2_fp32\n │ └── relayvsr_flash_vae_x4_fp32\n └── facexlib/\n ├── detection_mobilenet0.25_Final.pth\n └── parsing_bisenet.pth\n```", "src": "36dc9909", "srcTitle": "Relay Notes" }, "zh-TW": { "text": "## LC Relay VSR Upscale 🧩\n\n**這是什麼？**\n- 一步完成的影片和圖片放大器（2x 或 4x）。模型、VAE 和取樣器都已內建。速度快、清晰，還能加上真實的皮膚和頭髮細節。\n- 耗時只有 VOSR2 或 SeedVR2 的一小部分，品質更高。\n- **detect_faces** 會自動找到人臉，並壓住 Relay 在臉上的顆粒，讓皮膚保持原樣。\n- 💡 如果你用了細節修復（detailer），或想保護全部皮膚，把遮罩接到 **protect_mask**。它會覆蓋 detect_faces。\n\n- ✋ **第一次使用會下載模型：** 每種放大倍數 5.2 GB（bf16 為 10 GB），加一個 128 MB 的 VAE，人臉偵測另需 55 MB。開始時會跳出提示。按 Stop 可取消。已經有了？不論放在哪裡都找得到。\n\n---\n\n### 影片 🎥\n\n- **2x 開啟人臉** 是預設設定。畫面在動時，眼睛會直接盯著人臉，其餘部分都成了背景雜訊。\n - 只有在時間或檔案大小成為問題時，才使用 **關閉人臉**。\n- **4x** 能在 8 GB 顯示卡上執行，但在動態影片裡和 2x 幾乎看不出差別。多出來的銳利度反而可能讓畫面更難看。\n- 用 **LC Save Video** 的 **relay** 輸入儲存。它一次只存幾格，放大後的影片永遠不會整個放進記憶體（5 秒 4x：0.3 GB，而不是 12 GB）。\n - Relay 節點的 **image** 輸出保持空著，否則還是會產生完整影片。\n\n| 影片 | 2x 開啟人臉 | 2x 關閉人臉 | 4x 開啟人臉 | 4x 關閉人臉 |\n|---|---|---|---|---|\n| 960 x 544，5 秒，24 fps | 13 秒（估計） | 12 秒 | 36 秒（估計） | 34 秒 |\n| 896 x 1152，6.6 秒，24 fps | 26 秒（估計） | 22 - 24 秒 | 78 秒（估計） | 72 - 75 秒 |\n| 832 x 1056，15 秒，48 fps | 46 秒 | 40 秒 | 2.7 分鐘 | 2.6 分鐘 |\n| **8 GB 筆電：** 960 x 544，5 秒，24 fps | 28 秒（估計） | 27 秒 | 2 分鐘（估計） | 118 秒 |\n| **8 GB 筆電：** 832 x 1056，15 秒，48 fps | 4 分鐘 | 3.9 分鐘 | 18 分鐘（估計） | 17 分鐘（估計） |\n\n\\* 除註明外皆為 RTX 5090，筆電為 RTX 5060 8 GB。模型已下載。開啟人臉時每 3 格偵測一次人臉，多花約 4 - 15 %（在 15 秒片段上實測）。（估計）欄位由實測數據推算。\n- 💡 8 GB 顯示卡上顯存占用保持平穩（2x 時 7.4 GB），所以長片段不會越跑越慢。5090 在 2x 用了 13 GB、4x 用了 24 GB，只是因為它有空閒顯存。\n- ⚠️ **16 GB 記憶體：** 在 Relay 開始之前，所有畫格就已經放在記憶體裡了。15 秒、48 fps 的片段峰值為 18.5 GB。\n - 片段保持簡短，或者在補幀 **之前** 放大（畫格減半，記憶體減半）。建議做法，未測試。\n\n---\n\n### 圖片 🖼️\n\n- 大家真的會放大檢視圖片，所以這裡 4x 值得。圖片上開啟人臉幾乎不花時間（4x 時開關都約 4 秒，1056 x 1320 的圖片約 6 秒）。\n- **品質排序：**\n - 4x 開啟人臉\n - 2x 開啟人臉\n - 4x 關閉人臉\n - 2x 關閉人臉\n- **寫實和照片：** 開啟人臉。\n- **線稿、動漫、繪畫等：** 取決於臉。\n - 光滑或繪畫式的皮膚：開啟人臉（那裡的顆粒只是雜訊）。\n - 雀斑、排線或其他畫出來的細節：關閉人臉，否則這些細節會變軟。\n- 人群裡的小臉：開或關看不出差別。\n\n**💡 小技巧**\n- 用 **mask_protect_strength** 調整人臉受保護的程度。\n- 用 **detail** 調整銳利度。\n\n---\n\n### 設定 ⚙️\n\n- **source：**\n - **video：** 畫格之間互相匹配，影片保持穩定。凡是來自影片的內容都用它。\n - **image：** 每張圖片單獨放大，彼此之間不會互相影響。用於一批不同的圖片。單張圖片兩種方式結果相同。\n- **upscale：** 影片用 2x，圖片用 4x。每種都是獨立的模型。\n - **2x bf16 / 4x bf16（10 GB）：** 原始版本。可以選，但看不出明顯差別，還要多用約 5 GB 顯存。\n - 畫質：幾乎一樣。平均差異只有 0.12 - 0.21（滿分 255），100% 裁切圖也看不出差別。\n - 5090 上 4x：圖片都是 4 秒（顯存 10 - 11 GB 對 15 - 17 GB），3 秒影片 22 秒對 25 秒（19 GB 對 24 GB）。\n- **detail：** 1.0 是完整的模型輸出。調低會更接近一般放大。\n- **detect_faces：** 通常開啟。對帶雀斑或排線的手繪臉關閉。\n- **mask_protect_strength：** Relay 的細節還有多少會落到臉上。預設 0.5。調高可以在風格化的臉上保留更多紋理。\n\n---\n\n### 手動下載 📂（可選）\n\n- 節點會在第一次使用時下載這些檔案。不想在執行中途等待的話，可以自己先下載。\n\n- [Relay 2x 模型](https://huggingface.co/Comfy-Org/RelayVSR/resolve/main/diffusion_models/relayvsr_x2_lightx2v_int8_convrot.safetensors?download=true)\n\n- [Relay 2x VAE](https://huggingface.co/Comfy-Org/RelayVSR/resolve/main/vae/relayvsr_flash_vae_x2_fp32.safetensors?download=true)\n\n- [Relay 4x 模型](https://huggingface.co/Comfy-Org/RelayVSR/resolve/main/diffusion_models/relayvsr_x4_int8_convrot.safetensors?download=true)\n\n- [Relay 4x VAE](https://huggingface.co/Comfy-Org/RelayVSR/resolve/main/vae/relayvsr_flash_vae_x4_fp32.safetensors?download=true)\n\n- [Relay 4x bf16 模型（選用，10 GB）](https://huggingface.co/kopper/RelayVSR/resolve/main/wan.safetensors?download=true)\n\n- [Relay 2x bf16 模型（選用，10 GB）](https://huggingface.co/lightx2v/RelayVSR_X2/resolve/main/wan_x2.safetensors?download=true)\n\n```\nComfyUI/\n└── models/\n ├── diffusion_models/\n │ ├── relayvsr_x2_lightx2v_int8_convrot (5.2 GB)\n │ ├── relayvsr_x4_int8_convrot (5.2 GB)\n │ ├── wan_x2 (2x bf16, 選用, 10 GB)\n │ └── wan (4x bf16, 選用, 10 GB)\n ├── vae/\n │ ├── relayvsr_flash_vae_x2_fp32\n │ └── relayvsr_flash_vae_x4_fp32\n └── facexlib/\n ├── detection_mobilenet0.25_Final.pth\n └── parsing_bisenet.pth\n```", "src": "36dc9909", "srcTitle": "Relay Notes" }, "ja": { "text": "## LC Relay VSR Upscale 🧩\n\n**これは何？**\n- 動画と画像を1ステップで拡大するアップスケーラー（2x または 4x）。モデル、VAE、サンプラーはすべて内蔵。速くてシャープ、本物の肌と髪のディテールを加えます。\n- VOSR2 や SeedVR2 のほんの一部の時間で、しかも高品質。\n- **detect_faces** が顔を自動で見つけ、顔の上の Relay の粒状感を抑えるので、肌の見た目がそのまま残ります。\n- 💡 ディテーラーを使った場合や肌全体を守りたい場合は、マスクを **protect_mask** につなぎます。detect_faces より優先されます。\n\n- ✋ **初回使用時にモデルをダウンロードします：** 拡大サイズごとに 5.2 GB（bf16 は 10 GB）と 128 MB の VAE、顔検出用に 55 MB。開始するとトーストが出ます。Stop でキャンセル。もう持っている？ どこに置いてあっても見つかります。\n\n---\n\n### 動画 🎥\n\n- **2x 顔オン** がデフォルト。動いている映像では目はまっすぐ顔に向かい、それ以外は背景のノイズになります。\n - **顔オフ** は時間やファイルサイズが問題になるときだけ使います。\n- **4x** は 8 GB のカードでも動きますが、動いている映像では 2x とほとんど違いが分かりません。余分なシャープさでかえって悪く見えることもあります。\n- **LC Save Video** の **relay** 入力で保存します。数フレームずつ保存するので、拡大した動画が丸ごとメモリに乗ることはありません（5 秒の 4x：12 GB ではなく 0.3 GB）。\n - Relay ノードの **image** 出力は空けておきます。つなぐと結局フル動画が作られます。\n\n| 動画 | 2x 顔オン | 2x 顔オフ | 4x 顔オン | 4x 顔オフ |\n|---|---|---|---|---|\n| 960 x 544、5 秒、24 fps | 13 秒（推定） | 12 秒 | 36 秒（推定） | 34 秒 |\n| 896 x 1152、6.6 秒、24 fps | 26 秒（推定） | 22 - 24 秒 | 78 秒（推定） | 72 - 75 秒 |\n| 832 x 1056、15 秒、48 fps | 46 秒 | 40 秒 | 2.7 分 | 2.6 分 |\n| **8 GB ノートPC：** 960 x 544、5 秒、24 fps | 28 秒（推定） | 27 秒 | 2 分（推定） | 118 秒 |\n| **8 GB ノートPC：** 832 x 1056、15 秒、48 fps | 4 分 | 3.9 分 | 18 分（推定） | 17 分（推定） |\n\n\\* 注記がなければ RTX 5090、ノートPC は RTX 5060 8 GB。モデルはダウンロード済み。顔オンは 3 フレームごとに顔を検出し、約 4 - 15 % 余分にかかります（15 秒のクリップで実測）。（推定）は実測値から換算しています。\n- 💡 8 GB のカードでは VRAM が一定のまま（2x で 7.4 GB）なので、長いクリップでも極端に遅くなりません。5090 が 2x で 13 GB、4x で 24 GB 使ったのは、空きがあったからにすぎません。\n- ⚠️ **16 GB RAM：** Relay が始まる前に、フレームはすでにメモリに乗っています。15 秒、48 fps のクリップはピーク 18.5 GB。\n - クリップは短くするか、フレーム補間の **前に** 拡大します（フレームが半分、RAM も半分）。推奨、未テスト。\n\n---\n\n### 画像 🖼️\n\n- 画像は実際にズームして見られるので、ここでは 4x の価値があります。画像なら顔オンの時間はほぼゼロ（4x でオンでもオフでも約 4 秒、1056 x 1320 の画像で 6 秒）。\n- **品質の順番：**\n - 4x 顔オン\n - 2x 顔オン\n - 4x 顔オフ\n - 2x 顔オフ\n- **リアル系と写真：** 顔オン。\n- **線画、アニメ、ペイント系など：** 顔しだい。\n - なめらかな肌やペイントの肌：顔オン（そこでは粒状感はただのノイズ）。\n - そばかす、ハッチング、その他描き込まれたディテール：顔オフ。そうしないとそのディテールがぼやけます。\n- 群衆の中の小さな顔：オンでもオフでも見た目の違いはありません。\n\n**💡 コツ**\n- 顔の守り具合は **mask_protect_strength** で調整します。\n- シャープさは **detail** で調整します。\n\n---\n\n### 設定 ⚙️\n\n- **source：**\n - **video：** フレーム同士を合わせるので、動画が安定します。動画から来たものすべてに使います。\n - **image：** 画像ごとに個別に拡大するので、画像間で何も引き継がれません。別々の画像のバッチに使います。画像1枚ならどちらでも同じ結果です。\n- **upscale：** 動画は 2x、画像は 4x。それぞれ別のモデルです。\n - **2x bf16 / 4x bf16（10 GB）：** オリジナル版。選べますが、目に見える違いはなく、VRAM を約 5 GB 多く使います。\n - 画質：ほぼ同じ。平均差は 255 段階中 0.12 - 0.21 で、100% の切り抜きでも見分けがつきません。\n - 5090 で 4x：画像はどちらも 4 秒（VRAM 10 - 11 GB 対 15 - 17 GB）、3 秒のクリップは 22 秒対 25 秒（19 GB 対 24 GB）。\n- **detail：** 1.0 がモデルのフル出力。下げるほど普通の拡大に近づきます。\n- **detect_faces：** 通常はオン。そばかすやハッチングのある描かれた顔ではオフ。\n- **mask_protect_strength：** Relay のディテールが顔にどれだけ乗るか。デフォルトは 0.5。上げると様式化された顔のテクスチャがより残ります。\n\n---\n\n### 手動ダウンロード 📂（任意）\n\n- ノードは初回使用時にこれらをダウンロードします。実行中に待ちたくなければ、自分で先に入手してください。\n\n- [Relay 2x モデル](https://huggingface.co/Comfy-Org/RelayVSR/resolve/main/diffusion_models/relayvsr_x2_lightx2v_int8_convrot.safetensors?download=true)\n\n- [Relay 2x VAE](https://huggingface.co/Comfy-Org/RelayVSR/resolve/main/vae/relayvsr_flash_vae_x2_fp32.safetensors?download=true)\n\n- [Relay 4x モデル](https://huggingface.co/Comfy-Org/RelayVSR/resolve/main/diffusion_models/relayvsr_x4_int8_convrot.safetensors?download=true)\n\n- [Relay 4x VAE](https://huggingface.co/Comfy-Org/RelayVSR/resolve/main/vae/relayvsr_flash_vae_x4_fp32.safetensors?download=true)\n\n- [Relay 4x bf16 モデル（任意、10 GB）](https://huggingface.co/kopper/RelayVSR/resolve/main/wan.safetensors?download=true)\n\n- [Relay 2x bf16 モデル（任意、10 GB）](https://huggingface.co/lightx2v/RelayVSR_X2/resolve/main/wan_x2.safetensors?download=true)\n\n```\nComfyUI/\n└── models/\n ├── diffusion_models/\n │ ├── relayvsr_x2_lightx2v_int8_convrot (5.2 GB)\n │ ├── relayvsr_x4_int8_convrot (5.2 GB)\n │ ├── wan_x2 (2x bf16, 任意, 10 GB)\n │ └── wan (4x bf16, 任意, 10 GB)\n ├── vae/\n │ ├── relayvsr_flash_vae_x2_fp32\n │ └── relayvsr_flash_vae_x4_fp32\n └── facexlib/\n ├── detection_mobilenet0.25_Final.pth\n └── parsing_bisenet.pth\n```", "src": "36dc9909", "srcTitle": "Relay Notes" }, "ko": { "text": "## LC Relay VSR Upscale 🧩\n\n**이게 뭔가요?**\n- 영상과 사진을 한 번에 키우는 업스케일러(2x 또는 4x). 모델, VAE, 샘플러가 모두 내장되어 있습니다. 빠르고 선명하며, 실제 피부와 머리카락 디테일을 더합니다.\n- VOSR2나 SeedVR2에 걸리는 시간의 일부만 들고, 품질은 더 높습니다.\n- **detect_faces**가 얼굴을 스스로 찾아 얼굴 위의 Relay 그레인을 눌러 주므로, 피부가 원래 모습을 유지합니다.\n- 💡 디테일러를 썼거나 피부 전체를 보호하고 싶다면 마스크를 **protect_mask**에 연결하세요. detect_faces보다 우선합니다.\n\n- ✋ **첫 사용 시 모델을 다운로드합니다:** 업스케일 크기마다 5.2 GB(bf16은 10 GB)와 128 MB VAE, 얼굴 감지용 55 MB. 시작하면 토스트 알림이 뜹니다. Stop을 누르면 취소됩니다. 이미 있나요? 어디에 두었든 찾아냅니다.\n\n---\n\n### 영상 🎥\n\n- **2x 얼굴 켜기**가 기본입니다. 화면이 움직일 때 눈은 곧장 얼굴로 가고, 나머지는 배경 노이즈가 됩니다.\n - **얼굴 끄기**는 시간이나 파일 크기가 문제일 때만 쓰세요.\n- **4x**는 8 GB 카드에서도 돌아가지만, 움직이는 영상에서는 2x와 거의 차이가 없습니다. 오히려 과한 선명함 때문에 더 나빠 보일 수 있습니다.\n- **LC Save Video**의 **relay** 입력으로 저장하세요. 몇 프레임씩 저장하므로, 업스케일된 영상이 통째로 RAM에 올라가지 않습니다(5초 4x: 12 GB 대신 0.3 GB).\n - Relay 노드의 **image** 출력은 비워 두세요. 연결하면 결국 전체 영상이 만들어집니다.\n\n| 영상 | 2x 얼굴 켜기 | 2x 얼굴 끄기 | 4x 얼굴 켜기 | 4x 얼굴 끄기 |\n|---|---|---|---|---|\n| 960 x 544, 5초, 24 fps | 13초 (추정) | 12초 | 36초 (추정) | 34초 |\n| 896 x 1152, 6.6초, 24 fps | 26초 (추정) | 22 - 24초 | 78초 (추정) | 72 - 75초 |\n| 832 x 1056, 15초, 48 fps | 46초 | 40초 | 2.7분 | 2.6분 |\n| **8 GB 노트북:** 960 x 544, 5초, 24 fps | 28초 (추정) | 27초 | 2분 (추정) | 118초 |\n| **8 GB 노트북:** 832 x 1056, 15초, 48 fps | 4분 | 3.9분 | 18분 (추정) | 17분 (추정) |\n\n\\* 따로 적지 않으면 RTX 5090, 노트북은 RTX 5060 8 GB. 모델은 다운로드된 상태. 얼굴 켜기는 3프레임마다 얼굴을 찾으며 약 4 - 15 % 더 걸립니다(15초 클립에서 측정). (추정)은 측정값을 바탕으로 계산했습니다.\n- 💡 8 GB 카드에서는 VRAM이 일정하게 유지되므로(2x에서 7.4 GB), 긴 클립도 크게 느려지지 않습니다. 5090이 2x에서 13 GB, 4x에서 24 GB를 쓴 것은 여유가 있었기 때문일 뿐입니다.\n- ⚠️ **16 GB RAM:** Relay가 시작되기 전에 프레임이 이미 RAM에 올라가 있습니다. 15초, 48 fps 클립은 최고 18.5 GB였습니다.\n - 클립을 짧게 하거나, 프레임 보간 **전에** 업스케일하세요(프레임 절반, RAM 절반). 권장, 테스트 안 됨.\n\n---\n\n### 사진 🖼️\n\n- 사람들은 사진을 실제로 확대해서 보므로, 여기서는 4x가 그만한 값을 합니다. 사진에서는 얼굴 켜기에 드는 시간이 거의 없습니다(4x에서 켜든 끄든 약 4초, 1056 x 1320 사진은 6초).\n- **품질 순서:**\n - 4x 얼굴 켜기\n - 2x 얼굴 켜기\n - 4x 얼굴 끄기\n - 2x 얼굴 끄기\n- **실사와 사진:** 얼굴 켜기.\n- **라인아트, 애니메이션, 페인팅 등:** 얼굴에 따라 다릅니다.\n - 매끈하거나 칠한 피부: 얼굴 켜기(거기서는 그레인이 그냥 노이즈입니다).\n - 주근깨, 해칭 등 그려 넣은 디테일: 얼굴 끄기. 안 그러면 그 디테일이 뭉개집니다.\n- 군중 속 작은 얼굴: 켜든 끄든 눈에 띄는 차이가 없습니다.\n\n**💡 팁**\n- 얼굴을 얼마나 보호할지는 **mask_protect_strength**로 조절합니다.\n- 선명도는 **detail**로 조절합니다.\n\n---\n\n### 설정 ⚙️\n\n- **source:**\n - **video:** 프레임끼리 서로 맞추므로 영상이 안정적입니다. 영상에서 나온 것에는 모두 이것을 쓰세요.\n - **image:** 사진마다 따로 업스케일하므로 사진 사이에 아무것도 이어지지 않습니다. 서로 다른 사진 묶음에 쓰세요. 사진 한 장은 어느 쪽이든 결과가 같습니다.\n- **upscale:** 영상은 2x, 사진은 4x. 각각 별도의 모델입니다.\n - **2x bf16 / 4x bf16 (10 GB):** 원본 릴리스. 선택할 수 있지만 눈에 띄는 차이가 없고 VRAM을 약 5 GB 더 씁니다.\n - 화질: 사실상 같습니다. 평균 차이는 255단계 중 0.12 - 0.21이고, 100% 크롭으로도 구분이 안 됩니다.\n - 5090에서 4x: 사진은 둘 다 4초(VRAM 10 - 11 GB 대 15 - 17 GB), 3초 클립은 22초 대 25초(19 GB 대 24 GB).\n- **detail:** 1.0이 모델의 전체 출력입니다. 낮출수록 일반 업스케일에 가까워집니다.\n- **detect_faces:** 보통 켭니다. 주근깨나 해칭이 있는 그린 얼굴에서는 끕니다.\n- **mask_protect_strength:** Relay의 디테일이 얼굴에 얼마나 남는지. 기본값 0.5. 높이면 스타일화된 얼굴에 텍스처가 더 남습니다.\n\n---\n\n### 수동 다운로드 📂 (선택)\n\n- 노드가 첫 사용 시 이 파일들을 다운로드합니다. 실행 도중 기다리기 싫다면 직접 받아 두세요.\n\n- [Relay 2x 모델](https://huggingface.co/Comfy-Org/RelayVSR/resolve/main/diffusion_models/relayvsr_x2_lightx2v_int8_convrot.safetensors?download=true)\n\n- [Relay 2x VAE](https://huggingface.co/Comfy-Org/RelayVSR/resolve/main/vae/relayvsr_flash_vae_x2_fp32.safetensors?download=true)\n\n- [Relay 4x 모델](https://huggingface.co/Comfy-Org/RelayVSR/resolve/main/diffusion_models/relayvsr_x4_int8_convrot.safetensors?download=true)\n\n- [Relay 4x VAE](https://huggingface.co/Comfy-Org/RelayVSR/resolve/main/vae/relayvsr_flash_vae_x4_fp32.safetensors?download=true)\n\n- [Relay 4x bf16 모델 (선택, 10 GB)](https://huggingface.co/kopper/RelayVSR/resolve/main/wan.safetensors?download=true)\n\n- [Relay 2x bf16 모델 (선택, 10 GB)](https://huggingface.co/lightx2v/RelayVSR_X2/resolve/main/wan_x2.safetensors?download=true)\n\n```\nComfyUI/\n└── models/\n ├── diffusion_models/\n │ ├── relayvsr_x2_lightx2v_int8_convrot (5.2 GB)\n │ ├── relayvsr_x4_int8_convrot (5.2 GB)\n │ ├── wan_x2 (2x bf16, 선택, 10 GB)\n │ └── wan (4x bf16, 선택, 10 GB)\n ├── vae/\n │ ├── relayvsr_flash_vae_x2_fp32\n │ └── relayvsr_flash_vae_x4_fp32\n └── facexlib/\n ├── detection_mobilenet0.25_Final.pth\n └── parsing_bisenet.pth\n```", "src": "36dc9909", "srcTitle": "Relay Notes" } }, "titles": { "en": "Relay Notes", "zh": "中继笔记", "zh-TW": "中繼筆記", "ja": "Relay メモ", "ko": "Relay 노트" } }, "ue_properties": { "widget_ue_connectable": {}, "version": "7.8", "input_ue_unconnectable": {} } }, "widgets_values": [ "English ✏️", "## LC Relay VSR Upscale 🧩\n\n**What is it?**\n- A one-step upscaler for video and pictures (2x or 4x). The model, VAE and sampler are built in. Fast, sharp, and it adds real skin and hair detail.\n- A fraction of the time VOSR2 or SeedVR2 takes, and higher quality.\n- **detect_faces** finds the faces on its own and holds back Relay's grain on them, so skin keeps its look.\n- 💡 If you used detailers or want all skin protected, wire a mask into **protect_mask**. It overrides detect_faces.\n\n- ✋ **First use downloads the models:** 5.2 GB per upscale size (10 GB for bf16) plus a 128 MB VAE, and 55 MB for the face finder. A toast tells you it started. Press Stop to cancel. Already have them? They are found wherever they are.\n\n---\n\n### Videos 🎥\n\n- **2x faces on** is the default. Your eye goes straight to the faces when things are moving, and everything else is background noise.\n - Use **faces off** only when time or file size is the problem.\n- **4x** runs on an 8 GB card, but on a moving video it looks barely different from 2x. Extra sharpness can actually make it look worse.\n- Save with **LC Save Video**'s **relay** input. It saves a few frames at a time, so the upscaled video never sits in RAM (5 s at 4x: 0.3 GB instead of 12 GB).\n - Leave the Relay node's **image** output empty, or the full video gets built anyway.\n\n| Video | 2x faces on | 2x faces off | 4x faces on | 4x faces off |\n|---|---|---|---|---|\n| 960 x 544, 5 s, 24 fps | 13 s (est.) | 12 s | 36 s (est.) | 34 s |\n| 896 x 1152, 6.6 s, 24 fps | 26 s (est.) | 22 - 24 s | 78 s (est.) | 72 - 75 s |\n| 832 x 1056, 15 s, 48 fps | 46 s | 40 s | 2.7 min | 2.6 min |\n| **8 GB laptop:** 960 x 544, 5 s, 24 fps | 28 s (est.) | 27 s | 2 min (est.) | 118 s |\n| **8 GB laptop:** 832 x 1056, 15 s, 48 fps | 4 min | 3.9 min | 18 min (est.) | 17 min (est.) |\n\n\\* RTX 5090 unless noted, laptop is an RTX 5060 8 GB. Models already downloaded. Faces on finds the faces every 3rd frame and costs about 4 - 15 % (measured on the 15 s clip). (est.) rows are scaled from the measured ones.\n- 💡 VRAM stays flat on 8 GB cards (7.4 GB at 2x), so long clips do not slow to a crawl. The 5090 used 13 GB at 2x and 24 GB at 4x only because it had it free.\n- ⚠️ **16 GB RAM:** the frames sit in RAM before Relay even starts. The 15 s, 48 fps clip peaked at 18.5 GB.\n - Keep clips short, or upscale **before** frame interpolation (half the frames, half the RAM). Recommended, not tested.\n\n---\n\n### Pictures 🖼️\n\n- People actually zoom in on a picture, so 4x is worth it here. Faces cost nothing on a picture (about 4 s either way at 4x, 6 s for a 1056 x 1320 picture).\n- **Quality order:**\n - 4x faces on\n - 2x faces on\n - 4x faces off\n - 2x faces off\n- **Realism and photos:** faces on.\n- **Lineart, anime, painted, etc.:** depends on the face.\n - Smooth or painted skin: faces on (the grain is just noise there).\n - Freckles, hatching or other drawn detail: faces off, or that detail goes soft.\n- Small faces in a crowd: on or off makes no visible difference.\n\n**💡 Tips & tricks**\n- Adjust how much the faces are protected with **mask_protect_strength**.\n- Adjust the sharpness with **detail**.\n\n---\n\n### Settings ⚙️\n\n- **source:**\n - **video:** the frames are matched to each other, so the video stays steady. Use it for anything that came from a video.\n - **image:** every picture is upscaled on its own, so nothing carries over between them. Use it for a batch of different pictures. One picture comes out the same either way.\n- **upscale:** 2x for video, 4x for pictures. Each one is its own model.\n - **2x bf16 / 4x bf16 (10 GB):** the original release. It's an option, but not noticeably different, and it uses about 5 GB more VRAM.\n - Quality: practically identical. The average difference is 0.12 - 0.21 levels out of 255, and the 100% crops can't be told apart.\n - 4x on a 5090: pictures 4 s either way (10 - 11 GB vs 15 - 17 GB VRAM), a 3 s clip 22 s vs 25 s (19 GB vs 24 GB).\n- **detail:** 1.0 is the full model output. Lower moves toward a plain upscale.\n- **detect_faces:** typically on. Off for drawn faces with freckles or hatching.\n- **mask_protect_strength:** how much of Relay's detail still lands on the face. 0.5 is the default. Raise it to keep more texture on stylized faces.\n\n---\n\n### Manual download 📂 (optional)\n\n- The node downloads these on first use. Grab them yourself if you'd rather not wait mid-run.\n\n- [Relay 2x model](https://huggingface.co/Comfy-Org/RelayVSR/resolve/main/diffusion_models/relayvsr_x2_lightx2v_int8_convrot.safetensors?download=true)\n\n- [Relay 2x VAE](https://huggingface.co/Comfy-Org/RelayVSR/resolve/main/vae/relayvsr_flash_vae_x2_fp32.safetensors?download=true)\n\n- [Relay 4x model](https://huggingface.co/Comfy-Org/RelayVSR/resolve/main/diffusion_models/relayvsr_x4_int8_convrot.safetensors?download=true)\n\n- [Relay 4x VAE](https://huggingface.co/Comfy-Org/RelayVSR/resolve/main/vae/relayvsr_flash_vae_x4_fp32.safetensors?download=true)\n\n- [Relay 4x bf16 model (optional, 10 GB)](https://huggingface.co/kopper/RelayVSR/resolve/main/wan.safetensors?download=true)\n\n- [Relay 2x bf16 model (optional, 10 GB)](https://huggingface.co/lightx2v/RelayVSR_X2/resolve/main/wan_x2.safetensors?download=true)\n\n```\nComfyUI/\n└── models/\n ├── diffusion_models/\n │ ├── relayvsr_x2_lightx2v_int8_convrot (5.2 GB)\n │ ├── relayvsr_x4_int8_convrot (5.2 GB)\n │ ├── wan_x2 (2x bf16, optional, 10 GB)\n │ └── wan (4x bf16, optional, 10 GB)\n ├── vae/\n │ ├── relayvsr_flash_vae_x2_fp32\n │ └── relayvsr_flash_vae_x4_fp32\n └── facexlib/\n ├── detection_mobilenet0.25_Final.pth\n └── parsing_bisenet.pth\n```" ], "widgets_values_named": { "language": "English ✏️", "text": "## LC Relay VSR Upscale 🧩\n\n**What is it?**\n- A one-step upscaler for video and pictures (2x or 4x). The model, VAE and sampler are built in. Fast, sharp, and it adds real skin and hair detail.\n- A fraction of the time VOSR2 or SeedVR2 takes, and higher quality.\n- **detect_faces** finds the faces on its own and holds back Relay's grain on them, so skin keeps its look.\n- 💡 If you used detailers or want all skin protected, wire a mask into **protect_mask**. It overrides detect_faces.\n\n- ✋ **First use downloads the models:** 5.2 GB per upscale size (10 GB for bf16) plus a 128 MB VAE, and 55 MB for the face finder. A toast tells you it started. Press Stop to cancel. Already have them? They are found wherever they are.\n\n---\n\n### Videos 🎥\n\n- **2x faces on** is the default. Your eye goes straight to the faces when things are moving, and everything else is background noise.\n - Use **faces off** only when time or file size is the problem.\n- **4x** runs on an 8 GB card, but on a moving video it looks barely different from 2x. Extra sharpness can actually make it look worse.\n- Save with **LC Save Video**'s **relay** input. It saves a few frames at a time, so the upscaled video never sits in RAM (5 s at 4x: 0.3 GB instead of 12 GB).\n - Leave the Relay node's **image** output empty, or the full video gets built anyway.\n\n| Video | 2x faces on | 2x faces off | 4x faces on | 4x faces off |\n|---|---|---|---|---|\n| 960 x 544, 5 s, 24 fps | 13 s (est.) | 12 s | 36 s (est.) | 34 s |\n| 896 x 1152, 6.6 s, 24 fps | 26 s (est.) | 22 - 24 s | 78 s (est.) | 72 - 75 s |\n| 832 x 1056, 15 s, 48 fps | 46 s | 40 s | 2.7 min | 2.6 min |\n| **8 GB laptop:** 960 x 544, 5 s, 24 fps | 28 s (est.) | 27 s | 2 min (est.) | 118 s |\n| **8 GB laptop:** 832 x 1056, 15 s, 48 fps | 4 min | 3.9 min | 18 min (est.) | 17 min (est.) |\n\n\\* RTX 5090 unless noted, laptop is an RTX 5060 8 GB. Models already downloaded. Faces on finds the faces every 3rd frame and costs about 4 - 15 % (measured on the 15 s clip). (est.) rows are scaled from the measured ones.\n- 💡 VRAM stays flat on 8 GB cards (7.4 GB at 2x), so long clips do not slow to a crawl. The 5090 used 13 GB at 2x and 24 GB at 4x only because it had it free.\n- ⚠️ **16 GB RAM:** the frames sit in RAM before Relay even starts. The 15 s, 48 fps clip peaked at 18.5 GB.\n - Keep clips short, or upscale **before** frame interpolation (half the frames, half the RAM). Recommended, not tested.\n\n---\n\n### Pictures 🖼️\n\n- People actually zoom in on a picture, so 4x is worth it here. Faces cost nothing on a picture (about 4 s either way at 4x, 6 s for a 1056 x 1320 picture).\n- **Quality order:**\n - 4x faces on\n - 2x faces on\n - 4x faces off\n - 2x faces off\n- **Realism and photos:** faces on.\n- **Lineart, anime, painted, etc.:** depends on the face.\n - Smooth or painted skin: faces on (the grain is just noise there).\n - Freckles, hatching or other drawn detail: faces off, or that detail goes soft.\n- Small faces in a crowd: on or off makes no visible difference.\n\n**💡 Tips & tricks**\n- Adjust how much the faces are protected with **mask_protect_strength**.\n- Adjust the sharpness with **detail**.\n\n---\n\n### Settings ⚙️\n\n- **source:**\n - **video:** the frames are matched to each other, so the video stays steady. Use it for anything that came from a video.\n - **image:** every picture is upscaled on its own, so nothing carries over between them. Use it for a batch of different pictures. One picture comes out the same either way.\n- **upscale:** 2x for video, 4x for pictures. Each one is its own model.\n - **2x bf16 / 4x bf16 (10 GB):** the original release. It's an option, but not noticeably different, and it uses about 5 GB more VRAM.\n - Quality: practically identical. The average difference is 0.12 - 0.21 levels out of 255, and the 100% crops can't be told apart.\n - 4x on a 5090: pictures 4 s either way (10 - 11 GB vs 15 - 17 GB VRAM), a 3 s clip 22 s vs 25 s (19 GB vs 24 GB).\n- **detail:** 1.0 is the full model output. Lower moves toward a plain upscale.\n- **detect_faces:** typically on. Off for drawn faces with freckles or hatching.\n- **mask_protect_strength:** how much of Relay's detail still lands on the face. 0.5 is the default. Raise it to keep more texture on stylized faces.\n\n---\n\n### Manual download 📂 (optional)\n\n- The node downloads these on first use. Grab them yourself if you'd rather not wait mid-run.\n\n- [Relay 2x model](https://huggingface.co/Comfy-Org/RelayVSR/resolve/main/diffusion_models/relayvsr_x2_lightx2v_int8_convrot.safetensors?download=true)\n\n- [Relay 2x VAE](https://huggingface.co/Comfy-Org/RelayVSR/resolve/main/vae/relayvsr_flash_vae_x2_fp32.safetensors?download=true)\n\n- [Relay 4x model](https://huggingface.co/Comfy-Org/RelayVSR/resolve/main/diffusion_models/relayvsr_x4_int8_convrot.safetensors?download=true)\n\n- [Relay 4x VAE](https://huggingface.co/Comfy-Org/RelayVSR/resolve/main/vae/relayvsr_flash_vae_x4_fp32.safetensors?download=true)\n\n- [Relay 4x bf16 model (optional, 10 GB)](https://huggingface.co/kopper/RelayVSR/resolve/main/wan.safetensors?download=true)\n\n- [Relay 2x bf16 model (optional, 10 GB)](https://huggingface.co/lightx2v/RelayVSR_X2/resolve/main/wan_x2.safetensors?download=true)\n\n```\nComfyUI/\n└── models/\n ├── diffusion_models/\n │ ├── relayvsr_x2_lightx2v_int8_convrot (5.2 GB)\n │ ├── relayvsr_x4_int8_convrot (5.2 GB)\n │ ├── wan_x2 (2x bf16, optional, 10 GB)\n │ └── wan (4x bf16, optional, 10 GB)\n ├── vae/\n │ ├── relayvsr_flash_vae_x2_fp32\n │ └── relayvsr_flash_vae_x4_fp32\n └── facexlib/\n ├── detection_mobilenet0.25_Final.pth\n └── parsing_bisenet.pth\n```" }, "color": "#2E3A46", "bgcolor": "#2E3A46" } ], "links": [ [ 17, 20, 0, 21, 0, "INT" ], [ 34, 39, 0, 28, 0, "IMAGE" ], [ 52, 15, 0, 27, 0, "IMAGE" ], [ 53, 13, 1, 40, 0, "AUDIO" ], [ 54, 13, 2, 41, 0, "FLOAT" ], [ 55, 13, 0, 42, 0, "IMAGE" ], [ 59, 45, 0, 48, 0, "IMAGE" ], [ 60, 47, 0, 48, 1, "STRING" ], [ 67, 39, 0, 57, 0, "IMAGE" ], [ 70, 57, 0, 36, 0, "IMAGE" ], [ 71, 57, 0, 28, 1, "IMAGE" ], [ 72, 12, 0, 58, 0, "IMAGE" ], [ 79, 61, 0, 62, 0, "IMAGE" ], [ 80, 63, 0, 64, 0, "STRING" ], [ 81, 47, 0, 64, 1, "STRING" ], [ 82, 64, 0, 62, 1, "STRING" ], [ 84, 15, 0, 65, 0, "IMAGE" ], [ 85, 48, 0, 65, 1, "IMAGE" ], [ 86, 65, 0, 61, 0, "IMAGE" ], [ 95, 12, 0, 69, 0, "IMAGE" ], [ 96, 43, 0, 69, 1, "AUDIO" ], [ 97, 44, 0, 69, 5, "FLOAT" ], [ 94, 58, 1, 69, 4, "LC_RELAY" ], [ 99, 39, 0, 71, 0, "IMAGE" ], [ 100, 71, 1, 57, 1, "MASK" ], [ 101, 71, 1, 72, 0, "MASK" ], [ 68, 33, 0, 57, 2, "INT" ], [ 73, 6, 0, 58, 2, "INT" ], [ 98, 46, 0, 69, 6, "STRING" ] ], "groups": [ { "id": 1, "title": "#Relay Upscale Video", "bounding": [ 2380, -2380, 340, 330 ], "color": "#8AA", "flags": {} }, { "id": 6, "title": "Image:", "bounding": [ 1660, -3180, 2530, 670 ], "color": "#b58b2a", "flags": {} }, { "id": 4, "title": "Relay Upscale", "bounding": [ 2200, -3140, 1270, 620 ], "color": "#3f789e", "flags": {} }, { "id": 3, "title": "#Relay Upscale Image", "bounding": [ 2440, -3100, 630, 570 ], "color": "#8AA", "flags": {} }, { "id": 8, "title": "Head End", "bounding": [ 1190, -3180, 430, 670 ], "color": "#3f789e", "flags": {} }, { "id": 10, "title": "Test (Delete if you don't want it)", "bounding": [ 4240, -3170, 880, 720 ], "color": "#3f789e", "flags": {} }, { "id": 17, "title": "Video:", "bounding": [ 1660, -2470, 1700, 920 ], "color": "#A88", "flags": {} }, { "id": 16, "title": "Relay Video", "bounding": [ 2140, -2420, 1210, 860 ], "color": "#3f789e", "flags": {} } ], "config": {}, "extra": { "k2cf_state_id": "k2cf_1794218693220_m5ffhjq6fo", "ue_links": [], "vision_assistant": { "snapshots": [] }, "links_added_by_ue": [], "ds": { "scale": 1.0152559799477343, "offset": [ -727.7113254686565, 2243.8893645886237 ] }, "frontendVersion": "1.58.5", "VHS_latentpreview": false, "VHS_latentpreviewrate": 0, "VHS_MetadataImage": true, "VHS_KeepIntermediate": true }, "extensions": { "seed_widgets": { "57": 2, "58": 2 } }, "version": 0.4, "seed_widgets": { "57": 2, "58": 2 } }