Summary
Keywords
Full Transcript
In this episode of the ComfyUI tutorial series, you'll learn how to dramatically boost your image generation speed using Nunchaku with Flux Dev and Flux Kontext. This method brings real-time generation down to just 7 seconds on compatible RTX GPUs. We'll walk through a full installation process using a simplified installer (thanks to Ivo), which sets up a portable version of ComfyUI with all necessary nodes, Python, Git, and dependencies, without affecting your existing installation. You’ll also learn how to reuse models from your current setup to save space and time. ⚠️ System requirements: - Windows OS - NVIDIA RTX GPU What you'll learn: - How to install the Nunchaku-compatible version of ComfyUI - How to load and optimize Flux and Flux Kontext models - How to convert existing workflows to use Nunchaku - How to fix common GPU compatibility errors - How to use prebuilt workflows for generation, image editing, and product mockups - How to integrate Loras from CivitAI for style transfers and concept control - How to upscale quickly using optimized nodes and model settings Get the workflows and instructions from discord for free First use this link to join if you didn't join already https://discord.gg/gggpkVgBf3 Then you can access directly this episode using this link: https://discord.com/channels/1245221993746399232/1392157442187460699/1392157930958229604 Tip you can type # on discord and then nunchaku and press enter you can access that channel. Check Other Episodes https://www.youtube.com/playlist?list=PL-pohOSaL8P9kLZP8tQ1K1QWdZEgwiBM0 Unlock exclusive perks by joining our channel: https://www.youtube.com/channel/UCmMbwA-s3GZDKVzGZ-kPwaQ/join !!! The workflows works only with Nvidia Cards on Windows Easy ComfyUI installer with Nunchaku https://github.com/Tavris1/ComfyUI-Easy-Install/tree/Windows-Nunchaku update use this instead https://github.com/Tavris1/ComfyUI-Easy-Install and then install the bat wirh numchaku in name from addons ------------ For RTX 20,30,40 series use int4 version For RTX 50 series use fp4 version Get Flux Dev Nunchaku version from here https://huggingface.co/mit-han-lab/nunchaku-flux.1-dev/tree/main Get Flux Kontext Dev Nunchaku version from here https://huggingface.co/mit-han-lab/nunchaku-flux.1-kontext-dev/tree/main Place them to diffusion_models folder ..ComfyUI/models/diffusion_models ------------ First Clip: Download clip_l.safetensors from here: https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors?download=true Place it in the clip folder ..ComfyUI/models/clip Second Clip: Choose one of this models t5xxl_fp8_e4m3fn_scaled.safetensors from here https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp8_e4m3fn_scaled.safetensors or awq-int4-flux.1-t5xxl.safetensors from here https://huggingface.co/mit-han-lab/nunchaku-t5/resolve/main/awq-int4-flux.1-t5xxl.safetensors?download=true Place it in the clip folder. For some older video cards awq-int4-flux.1-t5xxl can give an error like this kernel is not supported on your gpu un that case use that version t5xxl_fp8_e4m3fn_scaled.safetensors ------------ Download ae.safetensor from here https://huggingface.co/Comfy-Org/Lumina_Image_2.0_Repackaged/resolve/main/split_files/vae/ae.safetensors?download=true place it in the VAE Folder ------------ Download 4x_NMKD-Siax_200k.pth mode from here https://huggingface.co/Akumetsu971/SD_Anime_Futuristic_Armor/resolve/main/4x_NMKD-Siax_200k.pth?download=true Place it in the upscale_models folder ------------ #fluxkontext #comfyuitutorial #nunchaku Chapters 00:00 Installing ComfyUI with Nunchaku 02:47 Extra Models Path - Get models from a different ComfyUI 03:58 Using optimized Flux Dev workflows + Errors 11:22 Flux Kontext Nunchaku workflows
Continue this lesson in the app
Install CourseHive on Android or iOS to keep learning while you move.
