Train and block edit and save LoRAs directly inside ComfyUI for Z-image, Flux Klein, SDXL, Flux, WAN 2.2, SD 1.5
A ComfyUI node pack that implements FreeLong (NeurIPS 2024) spectral blending for Wan 2.2 video generation
LoRA Pilot is an ultimate docker image for all Stable Diffusion LoRA trainers. Includes kohya_ss, diffusion pipes and TensorBoard for trainings and ComfyUI and InvokeAI for validation. Features shared models, modules, custom integrations and automatization scripts.
ComfyUI node for WAN I2V control with built-in person/face mask generation using MediaPipe
Non-linear video editing and generation based on Comfy backend.
Wan 2.2 Image-to-Video inference pipeline deployed on RunPod Serverless using a Docker-based GPU setup.
Docker service stack for a comprehensive local ai experience, running on a normal home hardware setup with single GPU. Project-Status: Working, Backlog.
ComfyUI_Wan22Blockswap tries to uzse advanced VRAM optimization techniques for WAN 2.1/2.2 models.
Timeline-based video editor for ComfyUI with scenes, project assets, gallery, images as keyframes, render staging, and takes for iterative long-form generation.
Hardware-adaptive text-to-video and image-to-video (Wan 2.1/2.2, LTX-Video). Probes the machine, then plans dtype, quantization and offloading to fit — from a 24 GB workstation down to an 8 GB consumer GPU or CPU-only. CLI, REST API and web UI.
Wan2.2 TI2V 5B image-to-video extension for Modly
16GB VRAM local AI video/image generation guide: real ComfyUI benchmarks of Wan2.2-5B / LTX-Video / HunyuanVideo + Z-Image Turbo, zero-dependency scripts, workflows and pitfalls | 16GB 显存本地 AI 视频/图片生成实战指南
Seven local-inference studies on one 32 GB Apple Silicon machine: what ships, what it costs, and how it fails.
Run ComfyUI with WAN 2.2 image-to-video workflows on Modal.
Vibe-coded, no warranty. Scan ComfyUI workflows for missing models and download them from HuggingFace or CivitAI into the correct folder. Smart folder routing, disk-space-saving filesystem links, live progress tracking.
Generate 768×1280 (81 frames, ~5s at 16fps) video from text prompts using the Wan 2.2 T2V-A14B Mixture-of-Experts diffusion model on a trn2.48xlarge instance with PyTorch Native (device='neuron').
To associate your repository with the wan22 topic, visit your repo's landing page and select "manage topics."