Nodes/Runware.ai ComfyUI Inference API Integration/ControlNet Preprocess Lineart Anime
ComfyUI Node

ControlNet Preprocess Lineart Anime

Clean Lines for Cleaner Characters

By Runware·Created 2 years ago·Updated about a month ago· 140
ControlNet Preprocess Lineart Anime
  • image
  • image
ttlfalse
ttl_value60
outputFormatJPG

General lineart preprocessing works fine on photos and clean illustration, but anime art has its own line language - thin, confident, stylized strokes - and a generic extractor tends to smear it into mush. Runware_controlnet_preprocess_lineart_anime is the variant tuned for that look: it pulls clean, anime-style lines out of a reference and hands them to your ControlNet chain, hosted on Runware's cloud. If your project lives anywhere in the manga/anime/illustration space, this is the preprocessor to reach for over the generic lineart node.

How it works

Image in, anime-flavored line map out as an IMAGE. The extraction is trained/biased toward the clean stroke style that anime models expect, so the resulting guide holds the character's structure and gesture without the visual noise a photo-oriented extractor leaves behind. From there it's the same play as every preprocessor in this pack: the map feeds RunwareBuild_controlNet's guideImage, the builder tunes how hard the model follows it, and the prompt supplies everything the lines don't.

The inputs

The thinnest schema in the family:

  • image (required) - your reference frame or sketch.
  • outputFormat (JPG/PNG/WEBP) and ttl (toggle + value) - delivery details.

Output is image - the cleaned anime line map. Preview it before spending generation credits; the whole point is a clean map, and if yours came out cluttered, that's your signal to fix the source.

Wiring it up

The standard three-node chain: preprocess → RunwareBuild_controlNet (guideImage) → model's controlNet socket. For character consistency across scenes - the classic use case the KB flags for lineart - this is the anime-specific version of that trick: the lines pin the character's shape, and the prompt can change everything else between shots.

Install and setup

Same as every node here: ComfyUI Manager → "Runware" → install → restart, or clone and pip install -r ComfyUI-Runware/requirements.txt. Light deps (runware-sdk, pillow, soundfile), API key via Settings → Runware API key, RUNWARE_API_KEY, or runware auth login, funded credits (the ~$20 minimum top-up is the standing complaint). Preprocessing is fractions of a cent per call.

Troubleshooting

  • Lines come out too faint or too thick - that's the extractor's flavor on your particular image. Try a higher-contrast source, or fall back to the general lineart node and compare - they're both a run away.
  • Output ignores the lines - raise weight in the ControlNet builder and confirm the map is actually feeding guideImage, not the model node.
  • Doesn't look like the reference's line style - the preprocessor extracts structure, it doesn't copy a given artist's stroke weight. Structure is what ControlNet needs anyway.

It's a one-input node that exists to make one thing better - anime characters that stay recognizably themselves across generations. If you've been fighting the generic lineart node on illustration work, this is the variant you were missing.

CategoryRunware/Image/runware

Inputs (4)

NameTypeDefaultDescription
imageIMAGE
ttloptBOOLEANfalseEnable to set ttl. Off uses the model's default.
ttl_valueoptINT60Time-to-live (TTL) in seconds for generated content. Only applies when `outputType` is `URL`.
outputFormatoptCOMBOJPGFile format for the generated image.

Outputs (1)

NameTypeDescription
imageIMAGE