<h3 id="minimax-h3">MiniMax H3</h3>This is an advanced workflow with numerous options. It is not necessarily designed for beginners (although I encourage you to try it). For a simpler version, try my one, Here: <a target="_blank" rel="ugc" href="https://civarchive.com/models/2600919/lonecats-simple-workflows-minimax-h3-krea-2-zit-klein9b-illustrious-flux-wan22-qwen-image-edit">https://civarchive.com/models/2600919/lonecats-simple-workflows-minimax-h3-krea-2-zit-klein9b-illustrious-flux-wan22-qwen-image-edit</a><hr />Instagram: <a target="_blank" rel="ugc" href="https://www.instagram.com/synth.studio.models/">https://www.instagram.com/synth.studio.models/</a>Buy me a☕ <a target="_blank" rel="ugc" href="https://ko-fi.com/lonecatone">https://ko-fi.com/lonecatone</a>This represents hundreds of hours of work that I give away for free. If you enjoy it, Please make sure to 👍like and post to my workflows. That is how you support me!!! Also, feel free to ⚡tip 😉<hr /><h3 id="uses-my-repo-https:github.comlonecatone23">Uses my Repo <a target="_blank" rel="ugc" href="https://github.com/lonecatone23">https://github.com/lonecatone23</a></h3>All LC Notes now translate to the following:English, 中文, 繁體中文, Русский, 日本語 ,한국어, Français, Español, عربي, Türkçe, Português (BR), فارسی, עברית, Italiano<h3 id="v9.0">V9.0</h3>Over the course of building my Showrunner workflow (not released yet, but there is a preview in the AV node pack), I tested a lot of different upscale methods 90% of us were doing it wrong. read about it <a target="_blank" rel="ugc" href="https://civarchive.com/articles/38145/upscaling-and-why-you-may-be-doing-it-wrong">HERE</a>Removed Ultimate upscale and actually get better product<h3 id="v7.0i-got-tired-of-substandard-llm-wrappers-so-i-built-my-own">V7.0 I got tired of substandard LLM wrappers, so I built my own</h3><ul><li>LC Vision <a target="_blank" rel="ugc" href="https://github.com/lonecatone23/ComfyUI_LC_Vision_nodes">https://github.com/lonecatone23/ComfyUI_LC_Vision_nodes</a><ul><li>Actually looks at 4 reference images and 1 video when writing the prompt</li><li>Automatically loads everything you need to get started. no endless troubleshooting</li><li>Automatically downloads it's own models, including Abliterated ones</li></ul></li></ul>Also Updated<ul><li>Ultimate Upscaler<ul><li>insane quality at a fraction of the time it takes for Latent upscalers or other means of upscale</li><li>VRAM friendly</li></ul></li><li>Other optimization tweaks</li><li>OCD stuff</li></ul><h3 id="v6.1">V6.1</h3><ul><li>apparantly had a wire crossed and forgot to hook up the prompt on Ref2VA after messing with a new LLM. Both are repaired.</li></ul><h3 id="v6.0">V6.0</h3><ul><li>Added a bunch of new audio and video nodes including a load audio node you can trim in real time:</li><li><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/ccce84f0-a3d1-4619-9635-08daaf18d68a/original=true/ccce84f0-a3d1-4619-9635-08daaf18d68a.jpeg" /></li><li>Added a optimized VAE that allows you to choose between regular and tiled to help with OOM issues:</li><li>Added an upscale suite that actually works at 1.0 for low VRAM quality🤔</li><li><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/b3c336b1-5f73-4b0d-9eee-40bece39ad94/original=true/b3c336b1-5f73-4b0d-9eee-40bece39ad94.jpeg" /></li><li>Added a whole audio group and expanded the video post processing for killer sound and image control</li><li><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/3224acd1-e77a-42b8-b31a-5d4d32b8bbf6/original=true/3224acd1-e77a-42b8-b31a-5d4d32b8bbf6.jpeg" /></li><li>Updated the save nodes. This is beta. I'm trying to get them to transfer data over to CivitAI.</li><li>Updated and added notes.</li></ul><h3 id="v5.0">V5.0</h3>I went back to the drawingboard. I borrowed an idea (or two) from @Plaguekind, researched several other optimization ideas, and made a change to the LLM:<ul><li>Swapped out cheesy LLM prompter for QwenVL with Minimax 🎞️ options.</li><li>Changed the entire setup on how the models swap.</li><li>Moved stuff around to confuse anyone who just got used to the other workflow.<ul><li>Eliminated duplicate loaders.</li><li>eliminated duplicate Preview windows.</li></ul></li><li>Swapped out the limited LoRA loaders for a Power LoRA loader.</li><li>Improved the Sampler setup.</li><li>Added and updated notes.</li><li>drank a lot of ☕</li></ul><h3 id="v4.0">V4.0</h3>I spent the last few weeks going through everyone else's workflows to see what worked and what did not. I built some custom nodes to deal with issues and tested out pretty much every optimization setup listed in the ComfyUI manager and on CivitAI I can get faster, but with less quality. I can go slower, but not gain much more quality. I tried chunking, Sparse attention, Optimizers, Cache options, etc. I trimmed it down to this:<ul><li>Overhauled the reference area</li><li>created pipes to better transfer data</li><li>Added Advanced Sparse Attention and H3 Optimizer nodes.<ul><li><a target="_blank" rel="ugc" href="https://github.com/Zironic/H3-Optimizations">https://github.com/Zironic/H3-Optimizations</a><ul><li>Speed is significantly faster with better quality than other setups.</li><li>even works with Sage Attention!!</li></ul></li></ul></li><li>Added <a target="_blank" rel="ugc" href="https://civarchive.com/models/2851079/h3-eros-max?modelVersionId=3275279">h3-eros-max </a>capabilities</li></ul><h3 id="v3.2">V3.2</h3>Man, I'm batting 1.000 as of late 🤔<ul><li>Added a second LoRA loader</li><li>Replaced the Sage attention nodes with the new H3 ones</li><li>Fixed the break there at the same time 🙄</li><li>made a connection redo in the reference area.</li></ul><h3 id="v2.0">V2.0</h3><ul><li>Added Turbo LoRAs</li><li>Added H3 Sigma schedulers</li><li>Massive update in notes</li><li>Added LoRA Loader that works for both models</li><li>Upgraded the LLM prompt Instructions to better align with the Model</li></ul>OverviewThis workflow includes a comprehensive set of features designed for advanced video generation and enhancement:<ul><li>Support for GGUF or standard diffusion models and CLIP models</li><li>Integrated LLM for intelligent prompt engineering</li><li>First-frame / last-frame to video (FLF2V) generation</li><li>Reference-to-video (REF2V) capability</li><li>Supports up to 4 reference images (expandable via additional connections)</li><li>Accepts video inputs</li><li>Accepts audio inputs</li><li>RIFE frame interpolation</li><li>RTX Super Resolution upscaling</li><li>也有中文说明</li></ul><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/43abbfcd-7342-40d5-8ae4-810d7f3ef22a/original=true/43abbfcd-7342-40d5-8ae4-810d7f3ef22a.jpeg" />

Lonecat’s MiniMax H3 FL2VA/ REF2VA | 低显存视频工作流 w/ Low VRAM Optimizers, Upscaler, & Post Production

V9.0 Over the course of building my Showrunner workflow (not released yet, but there is a preview in the AV node pack), I tested a lot of different upscale methods 90% of us were doing it wrong. read about it <a target="_blank" rel="ugc" href="https://civitai.red/articles/38145/upscaling-and-why-you-may-be-doing-it-wrong">HERE</a>Removed Ultimate upscale and actually get better product

MiniMax H3 Ver 9.0 (1).json

V9.0 Faster & Better

V8.0.1 Overhaul

V7.0.1 LC Vision

V6.1 WIre Fix

V6.0.1.0.2 Updates

V6.0 Pro Grade

V5.2 Minor Tweaks

V5.1 Overhaul & Optimize

V4.0 Overhaul

V3.2.1 Tweaks & Fix

V2.0 Turbo & Upgrades

v1.0 Beta

Files with SHA256: 2ac9ab9ff8ade81250b7e5e3ea7f0643530b1b14f09d5f2e799f44dd29d3773b