ComfyPilotComfyPilot
模型库 / LoRA / Near / Far — Subject Zoom Slider
Near / Far — Subject Zoom Slider

Near / Far — Subject Zoom Slider

3060559

LoRA 9 次下载 更新于 2026-11-22

底模:SDXL 1.0

conceptperspectivefull bodyportraitexperimentalcinematicphotographycomposition

下载模型 在源站打开

文件清单 线路:原始(huggingface.co) · 切换

下载后放进对应的 ComfyUI 目录。models/ 指 ComfyUI 的模型根目录。

权重 NF_Attempt_v1.safetensors 177 MB 下载

说明

Near / Far — Subject Zoom Slider

作者:DetroitWobbly

Near / Far — Subject Zoom Slider Get closer. Back away. Sometimes the shoes come with you. Near / Far is a bidirectional SDXL concept slider that changes the apparent distance between the viewer and the subject, allowing you to move from wider compositions toward intimate close-ups—or back again—without constantly rewriting your prompt. It works with people, animals, architecture, and other subjects. It can even follow an action rather than simply zooming toward someone's face. A woman tying her shoes, for example, may progress from a full-body composition toward a close-up of her hands and shoelaces. A tower can move from a distant architectural view toward a detailed crop of its stonework. Positive strength (+): Moves closer to the subject, producing tighter framing and close-ups. Negative strength (−): Moves farther away, producing wider views and more visible subject context. Recommended strengths

  • −3 to −2: Wider framing, depending on the prompt.
  • −1 to +1: Moderate framing adjustments.
  • +2: Strong close-up effect; watch for changes in identity or composition.
  • +3: Extreme close-ups and detail views. Results may become adventurous.

These are guidelines, not hard limits. The available range depends heavily on the prompt and checkpoint. Prompting: Give it room to move! Near / Far works best when your prompt describes what the subject is without unnecessarily dictating everything that must remain visible. For example: Character portrait A full-body portrait of a 35-year-old woman with auburn hair and green eyes, natural skin texture, realistic detailed materials, soft photographic lighting. This is a useful starting point for a wide framing sweep. The explicit full-body instruction helps establish the distant endpoint, while positive slider strengths can still push toward an extreme facial close-up. Animal portrait A full-body portrait of a chestnut horse, natural coat texture, detailed mane, realistic anatomy, soft daylight. Architecture An ancient stone tower, weathered masonry, realistic stone textures, natural atmospheric lighting. Action-focused composition A full-body portrait of an elderly watchmaker carefully repairing a mechanical pocket watch, realistic materials, intricate mechanical details, warm workshop lighting. With action prompts, the slider may move toward the activity itself—the watch, hands, or tools—rather than the person's face. A word about shoes The more specific your prompt becomes about visible clothing, props, actions, and surroundings, the less freedom the slider may have. For example: A woman wearing a green skirt and red high-heeled shoes. At wider framing, no problem. At closer framing, however, the checkpoint may desperately try to keep those shoes in the picture. In testing, one woman changed her pose to preserve her outfit, then brought a red high heel right up beside her face rather than allow it to disappear. The slider wanted a close-up. The prompt demanded shoes. SDXL negotiated. For greater framing freedom, try: A stylish woman elegantly dressed in green and red. General clothing descriptions usually work better than naming individual garments that the model feels compelled to display. The same principle applies to swords, handbags, horses in paddocks, and buildings on hills. Important limitations Near and Far are not perfectly symmetrical. Positive strengths can often override explicit full-body instructions. Negative strengths are much less effective against prompts that explicitly demand an extreme facial close-up. If you want to zoom outward, avoid close-up-specific language. The slider does not preserve a fixed scene or character identity like a physical camera moving through a 3D environment. It can change pose, perspective, clothing, facial details, object geometry, or background composition as it adjusts apparent proximity. At stronger settings, identity drift and unexpected reinterpretations become more likely, particularly with character LoRAs or highly constrained prompts. For best results, use a consistent seed, start with moderate strengths, and adjust the wording when the composition becomes locked. Tested behavior Evaluated on Juggernaut XL Ragnarok using a controlled 72-image prompt-constraint probe covering explicit framing, clothing, environments, props, actions, and architecture. All twelve tested strength ladders moved consistently in the intended framing direction, although several showed plateaus, abrupt composition changes, or identity drift. An additional 12-image test confirmed that explicit extreme-close-up language strongly restricts the negative/Far direction. Results may vary with other SDXL checkpoints and LoRA combinations.

示例图