SCAIL-2
#13 in KI-Video-EditingUnknown · v2 · 2× · tolest 21. Juli 2026
30
Momentum
SCAIL-2 is an Open Source model for end-to-end character animation developed by Z.ai (Zhipu AI) in collaboration with Tsinghua University. It transfers movements from a driver video directly onto a reference character image without requiring intermediate steps such as skeleton or pose extraction, and also supports character swapping and multi-person scenes. The model is based on Wan2.1 (14B) and was released simultaneously on arXiv, Hugging Face, and GitHub on June 9, 2026. In addition to the open model version, commercial API providers (e.g., fal.ai, WaveSpeedAI) exist, as well as separate end-user web platforms with credit systems.
Momentum-Verloop
08.05.06.08.
Features
| Output Formats | Video, 512p and 704p (or 480p/720p depending on provider), dimensions divisible by 32 |
| Base Model | Wan2.1 14B (video diffusion model) as backbone |
| Integrations | ComfyUI integration available; multi-reference support since June 2026 |
| Collaboration | No team/collaboration feature documented; web platforms offer job history and dashboard for individual users |
| License | Apache License 2.0 (commercial use permitted) |
| Platform | Local inference via GitHub code/ComfyUI; also available via cloud APIs (fal.ai, WaveSpeedAI) and web platforms |
| Price | Model weights free (open source); API usage e.g. via fal.ai/WaveSpeedAI from $0.20 per generated second |
| Release Date | June 9, 2026 |