🎞️SVD Styler

Remember the style dropdown on the old stablevideo.com site? 🎞️SVD Styler is that dropdown, ported into ComfyUI. Pick one of 17 presets - Analog Film, Cinematic, Pixel Art, and friends - and it wraps your prompt in the matching template and hands back a ready-to-use positive and negative prompt. It's from the ComfyUI-SVD-ZHO pack, ZHO's WIP recreation of that web app's interaction as nodes, and it's the node you'd reach for when you want a consistent "look" without typing the same style boilerplate into every prompt.

How it works

It's pure string surgery - no model, no CLIP, no API call, instant. The pack ships a style_template.py file with one template per style, and apply_style does two things: it swaps {prompt} in the template for your text, and it appends the style's negative prompt to the negative you typed. The output is just text, ready to feed into a CLIP Text Encode.

Inputs

• prompt - your actual subject. Multiline, with a default about a steaming cappuccino.
• negative_prompt - multiline too, defaults to asymmetry, worst quality, low quality.
• style_name - the enum with 18 choices: None plus 3D-Model, Analog Film, Anime, Cinematic, Comic Book, Digital Art, Enhance, Fantasy Art, Isometric, Line Art, Low Poly, Craft Clay, Neon Punk, Origami, Photographic, Pixel Art, and Tile Texture.

Outputs

Two STRINGs, named positive_prompt and negative_prompt. Wire positive_prompt into your positive CLIP Text Encode and negative_prompt into the negative one. That's the whole integration - it slots into any graph that has a text encoder.

Install

Same pack as the other two SVD helpers. ComfyUI Manager → search ComfyUI-SVD-ZHO → install, or:

cd ComfyUI/custom_nodes
git clone https://github.com/ZHO-ZHO-ZHO/ComfyUI-SVD-ZHO

then restart ComfyUI. No dependencies, no model files - the styles are hardcoded in the repo. You'll find it under Zho模块组/🎞️SVD.

Gotchas

One quirk worth knowing before it confuses you: the style's negative prompt always gets appended to yours, and that includes None, whose negative is literally ugly. So even with no style selected, the negative you get out isn't just what you typed. If your results suddenly look over-restricted, check what the node actually emitted. Also keep your expectations realistic: these templates were written for the web app's SVD 1.1, and they're generic prompt seasoning ("award-winning, professional, highly detailed") rather than anything model-specific - they work on whatever checkpoint you're using, but don't expect them to magically nail a style. And since the pack is WIP at v0.9 with SVD itself long in the tooth by 2026, treat this as a tidy convenience layer for a specific old workflow - one click instead of typing style boilerplate, nothing more, nothing less.