wan2

[ICML 2026] Official codebase for "Causal Forcing: Autoregressive Diffusion Distillation Done Right for High-Quality Real-Time Interactive Video Generation" & Causal Forcing++

A tiny graphical application for local AI generation

Mixed-precision quantization scheme (16/8/4bit mixed quantization) for the Wan2.2-Animate-14B model. Compresses the original 35GB base model to 17GB, balancing inference performance and model size.

WAN 2.x inference with ComfyUI

Local video and image generation queue for Apple Silicon that holds jobs instead of swapping

An unofficial, high-performance clone of the Wan 2.2 T2V architecture built in pure Python. Supports ComfyUI custom nodes, Flow Matching sampling, Causal 3D VAE, and cloud inference via Veo 3.1 API. Runs on any GPU without Docker or root access.

Turn any scene into a 1320×2868 iPhone wallpaper rendered as a late-Edo nishiki-e woodblock print — in the idiom of Yoshitoshi, Utamaro, Hokusai, or Kuniyoshi. Cloudflare Workers + Durable Object + Dashscope wan2.7-image-pro.

Build PackForcing for long video sampling and long-context inference with short video training and benchmark code

Wan2.2-TI2V-5B × ComfyUI 一鍵安裝包 (Ubuntu 22.04/24.04/26.04 + NVIDIA RTX,x86_64/aarch64),全程冪等,75 個 HF 模型 URL 已驗證

The first frame, and everything after it. A pipeline that turns a written scenario into a finished illustrated training film: keyframes, motion, narration, captions, cut and acceptance.

Identity-preserving image-to-video generation: vision-grounded prompt simplification via Qwen3-VL, Lightning LoRA 4-step inference, and SAM3-masked DINOv3 candidate reranking for fluid 720p video from a single reference image.

Self-hosted AI image and video generation for AMD Ryzen AI APUs on Linux: Vulkan (Mesa RADV) via stable-diffusion.cpp, no ROCm. Web UI with queue, model manager, multi-user auth and system diagnostics.

Local AI video generation — Wan 2.2 GGUF + LTX-Video via ComfyUI API. RTX 4090 optimized. No cloud.