Train a real-time, playable Video World Models on 8 GPUs — keyboard & mouse control & gamepad, fully open and reproducible.
Quartz — custom CUDA/Vulkan GGUF inference engine + SD1.5/SDXL image and video generation (Rust, no llama.cpp)
Local AI video generator: text-to-video on your own GPU with Wan2.1, Kokoro TTS and Whisper — offline, open source, no API keys.
Official code, locked protocols, and audit artifacts for "Jev-Routed Temporal Granularity with Recovery and Hysteresis for Causal Video Generation": per-segment fine/coarse/hold routing for Causal-rCM (Wan2.1-1.3B) via the Jev oracle, with two-GPU (T4/A100) feasibility audits.
a careful thought out approach to installing ComfyUI with Wan2.1, manager and some other specialty nodes that are a must on day one for me.
AI video generation browser for Wan2GP / WanGP: local library, full generation recipe per clip (seed, steps, LoRAs, prompt), settings diff and side-by-side / wipe / blink A/B comparison. Also browses ComfyUI output folders. 1-click Pinokio install, zero dependencies.
Real-time AI video generator: endless text-to-video you steer live with prompts. Runs locally on a single RTX 4090 with Wan 2.1 and Causal Forcing.
Unofficial ComfyUI integration of Causal Forcing / Causal Forcing++ (official frame-wise 2-step and 4-step checkpoints, official inference path, separate WSL2/Linux runtime)
ICDepth: Taming Video Diffusion Models for Video Depth Estimation via In-Context Conditioning (ECCV 2026)
CLI-first AI video agent: Wan 2.1/2.2 text-to-video, image-to-video and Wan Animate character replacement on on-demand A100 GPUs.
🎬 全链路智能短视频创作工作流 (CSCA 世界模型 + ComfyUI 6大岛链 + 15-Module 提示词工程 + Manim 竖屏双层合成器)
Adaptive ADHD task breakdown with on-demand AI video guides: Gemini 2.5 Flash + locally hosted Wan 2.1 (4-bit) with a custom LoRA, running on a 6 GB GPU
InfernoBench: open-source AI model installers, local inference benchmarks, and reproducible browser labs. Wan2.1 and Sulphur-2 included.
InfiniteTalk (Wan2.1) talking-avatar pipeline: ComfyUI workflows, RunPod pod/serverless setup, A/B matrix and forensic evaluation vs HeyGen
Unofficial ComfyUI nodes for DuoMatching streaming text-to-video (Wan2.1-T2V-1.3B, 2 steps per frame); in-process, no flash-attn; bit-identical to the official script
alsrory.AI - free, open-source, local-first AI video studio. Arabic RTL UI, Next.js + FastAPI, pluggable ComfyUI / Colab engine.
Production diagnostics, analytical VRAM budgeting, ComfyUI/CLI compiler, and Model Context Protocol (FastMCP) bridge for Wan-Video (Wan2.1 & Wan2.2 MoE).
Streaming video diffusion on one AWS Trainium2 chip. StreamDiffusionV2’s causal Wan2.1 1.3B model ported to Neuron with a resident KV cache and a 4-NeuronCore layer pipeline: 5.6 → 38.4 FPS end to end, matching the CPU reference (cosine 0.99996). Plus a live SD-Turbo webcam restyle at 43.8 FPS.
Make AI trailers 100% locally on a 6 GB GPU: video (Wan 2.1), music (ACE-Step), narration (OmniVoice) and ffmpeg mix, driven by a storyboard.json, by hand or by an LLM agent.
To associate your repository with the wan2-1 topic, visit your repo's landing page and select "manage topics."