Bonus Prompt Video Presets (Legacy)

Video models are even more demanding than image models about how you phrase things, and they're picky about what a style block is allowed to say. Bonus Prompt Video Presets (Legacy) is the video-flavored twin of Bonus Prompt Presets: the same style dropdown, but the returned paragraph is rewritten for video conditioning - an instruction, not a description.

Mechanically it's the image preset, passed through a prompt-rewrite step. Pick a style and the node looks up the base bonus prompt, then runs it through a "to-video-prompt" transform that converts image-editing language into role-aware video guidance and appends the appropriate video-style guidance block for the "bonus" role. The result is still a plain string, but its phrasing is tuned for how video text encoders read instructions - stable subject language, camera/style framing, temporal consistency hints - instead of the static-scene vocabulary the image version uses.

The inputs that matter

• preset - same style shelf as the image version: STYLE_3D, STYLE_ANIME, STYLE_PHOTOREALISM, the whole lot. Same names, video-flavored output.
• escape_parentheses - same guard as the image version; off by default.

The output

bonus_prompt - a STRING to append to your video prompt. Because the transform assumes the text is additional guidance, the phrasing leans on keeping the scene stable and coherent, which is exactly the role a bonus block should play in a video prompt. Wire it into your prompt concatenation ahead of the video text encoder.

Installing it

Part of silveroxides/ComfyUI-UtilsCollection. ComfyUI Manager → search ComfyUI-UtilsCollection → install → restart, or:

cd ComfyUI/custom_nodes
git clone https://github.com/silveroxides/ComfyUI-UtilsCollection

Restart after. Dependencies: opencv-python, typing-extensions. The preset text lives in the pack; no downloads. "(Legacy)" marks the compat alias for the pack's canonical UC_BonusPromptVideoPresets.

Where people get burned

The same non-negotiable as the image version: this is a bonus block, not a script. A video prompt that is only a style paragraph has no subject, no action, and no motion - you'll get a beautiful static nothing. Pair it with a real subject/action prompt. And if you swap between this and the image preset on the same workflow, double-check which one is wired - the phrasing difference is real, and using image-style bonus text on a video model is how you get a "style" that ignores the temporal structure entirely.