ToonCrafter
ToonCrafter is a free, open-source research project that interpolates two cartoon images into intermediate animation frames.
What ToonCrafter is
ToonCrafter is a free, open-source research project that interpolates two cartoon images into intermediate animation frames. It uses pretrained image-to-video diffusion priors, while sparse sketch guidance can influence the generated frames. The project also demonstrates cartoon sketch interpolation and reference-based sketch colorization with single or dual references. Local use is available through command-line inference and a Gradio demo, with links to a Hugging Face demo and community implementations for ComfyUI, Colab, and Windows. The model supports up to 16 frames at 512×320 resolution. The listed model uses about 24 GB of GPU memory, although community changes reportedly reduce this to about 10 GB; 50-step inference is reported at about 24 seconds on an A100. The makers describe ToonCrafter as a research exploration, not a commercial product, and say success is not guaranteed. They caution that image contents may be misunderstood and transitions can be unconvincing when objects appear or disappear. The repository lists an Apache-2.0 license.
Good
- Free and open source
- Supports sketch-guided interpolation
- Includes local command-line and Gradio use
- Links ComfyUI and Colab implementations
- Apache-2.0 license
Watch out
- Supports up to 16 frames
- Requires substantial GPU memory
- Generation success is not guaranteed
- Transitions may fail around appearing objects
Who it suits
It suits developers and researchers exploring cartoon interpolation, sketch guidance, and colorization through self-hosted or community-supported setups. Users should be comfortable with the stated hardware and model limitations.
Verdict
ToonCrafter provides an accessible research codebase for cartoon interpolation and colorization. Its frame, hardware, and reliability limits make it better suited to experimentation than dependable production output.
The facts
Each one read from ToonCrafter's own pages, with the date it was checked.
What it does
- Access
- The makers say they have not set up official profit-making projects or web applications.checked 4 Oct 2026
- Additional uses
- The project demonstrates cartoon sketch interpolation and reference-based sketch colorization.checked 4 Oct 2026
- Colorization
- The project demonstrates reference-based sketch colorization using single or dual image references.checked 9 Oct 2026
- Demo
- The project page links to a Hugging Face demo.checked 9 Oct 2026
- Hardware
- The repository reports about 24 GB GPU memory and 24 seconds per frame on an A100 for its listed 50-step configuration, and notes community reductions to about 10 GB.checked 9 Oct 2026
- Integrations
- The repository lists community implementations or support for ComfyUI, Colab, and Windows.checked 9 Oct 2026
- Known limitations
- The model may misinterpret image content and may struggle with transitions when objects appear or disappear.checked 9 Oct 2026
- License
- The GitHub repository identifies its code license as Apache-2.0.checked 9 Oct 2026
- Local use
- The repository provides command-line inference and a local Gradio demo.checked 4 Oct 2026
- Model limit
- The makers report support for videos up to 16 frames at 512×320 resolution.checked 4 Oct 2026
- Model limits
- The repository says generation supports up to 16 frames at 512x320 resolution.checked 9 Oct 2026
- Project status
- The authors describe ToonCrafter as an open-source research exploration rather than a commercial product and say no official profit-making project or web application has been set up.checked 9 Oct 2026
- Purpose
- ToonCrafter interpolates two cartoon images using pretrained image-to-video diffusion priors.checked 9 Oct 2026
- Research affiliation
- The project page lists contributors affiliated with the Chinese University of Hong Kong, City University of Hong Kong, Tencent AI Lab, and Monash University.checked 4 Oct 2026
- Research status
- The makers describe ToonCrafter as an open-source research exploration rather than a commercial product and say generation success is not guaranteed.checked 4 Oct 2026
- Sketch guidance
- It supports sparse sketch guidance for generated cartoon interpolation.checked 9 Oct 2026
- Ways to run
- The repository provides command-line inference and a local Gradio demo.checked 9 Oct 2026
The company
- Founded
- 2024checked 28 Sep 2026
