VintageAnima
ComfyUI node that builds a VintageAnima caption from series, cast, designer, scan, and quality.
VintageAnima for ComfyUI
A ComfyUI node that builds a VintageAnima caption in the same order as the training stills.
anime screenshot, year, decade, characters, series, description, hires or lowres, character designer, quality
The node is VintageAnima Prompt. It has two string outputs, positive and negative, ready to connect to CLIP Text Encode.
What you can set
- Prefix, series, year, and decade
- Up to three characters. Limit characters to this series is on by default, so those menus show the cast of the title you picked. Turn it off to choose someone from another series. Names you already picked stay in the caption.
- Character designer. From series uses the designer credited on that title.
- A large description box in the middle, for tags or a sentence
- Scan size:
hires,lowres, the scan that title was trained on, or omit - Quality tags
- A default negative, plus any extra negative text
Shot, expression, hair, clothing, and the other descriptive tags are left out on purpose. Those belong in the description, written the way you want them.
Empty slots are dropped. The description is lowercased and the spacing around commas is cleaned up.
Appleseed and Nausicaä of the Valley of the Wind each appear twice, because the two scans were trained differently. The label in parentheses tells them apart.
Default negative
Turn Default negative on and the negative takes the quality tags from the other scan:
- A hires caption negatives the lowres-only tags: blurry, low contrast, aliasing, telecine, dirty scan, ringing artifacts, vhs, rainbowing, composite artifacts, dot crawl.
- A lowres caption negatives the hires-only tags: hq cel animation, clean lineart.
Tags you already selected on the positive are left off the negative. Film grain, dust and scratches, soft focus, oversharpened, color banding, and faded colors were used on both scans, so they are not added automatically. Extra negative text is appended. Turn the default off and the negative is only that extra text.
Hires stills were trained with hq cel animation. Add it from the quality list when you want that look.
Install
cd ComfyUI/custom_nodes
git clone https://github.com/zetaneko/vintageanime-comfyui
Restart ComfyUI. The node is under Add Node → VintageAnima → VintageAnima Prompt.
The same caption builder is on the web at https://zetaneko.github.io/vintage-anima/