<h1 id="lazy-minimax-h3-all-in-one-(auto-switching)">Lazy MiniMax H3 — All-in-One (Auto-Switching)</h1>Pick the workflow type. Everything else follows.This is a Lazy MiniMax H3 graph: one selector, one conditioner path, one sampler stack. Set T2V / I2V / FL2V / R2V and the all-in-one nodes route models, frames, refs, and audio for you — no muted branches, no duplicate samplers, no “which UNET did I leave on?”<hr /><h2 id="the-lazy-idea">The Lazy idea</h2>Lazy here means the workflow does the bookkeeping.<ul><li>You choose the workflow type once.</li><li>Lazy MiniMax All-in-One builds the right conditioning for that mode.</li><li>Lazy Model Switcher picks fl2va (text / image / first–last) or ref2va (reference).</li><li>Lazy Image Loaders only emit what that mode needs; unused roles hard-gate off so optional sockets stay empty.</li><li>LazyPrompt can expand a short idea into a MiniMax-ready prompt — then get out of the way.</li></ul>Everything is modular. Nodes are grouped for clarity, not locked layout. Drag sections wherever you like; the wiring stays intentional.<hr /><h2 id="modes-at-a-glance">Modes at a glance</h2>ModeWhat you loadDiffusionT2VText onlyfl2vaI2VFirst framefl2vaFL2VFirst + last framefl2vaR2VReference images (+ optional audio)ref2vaChange the selector → loaders, MiniMax, and UNET routing update together.<hr /><h2 id="prompt-enhancement-(optional-lazy-friendly)">Prompt enhancement (optional, Lazy-friendly)</h2>You can run raw text straight through (Prompt Engineer bypass) or enhance first.<h3 id="recommended:-lm-studio-(local-llm)">Recommended: LM Studio (local LLM)</h3><ul><li>Workflow can load the model, enhance, then unload when finished.</li><li>Typical enhance time: about 5–10 seconds.</li><li>Best quality/speed balance for day-to-day Lazy runs.</li><li>Vision-capable LM Studio models work well with I2V first-frame grounding.</li></ul><h3 id="alternative:-textgenerate-(clip)">Alternative: TextGenerate (CLIP)</h3><ul><li>Uses a wired Comfy CLIP / Qwen LLM — no second app.</li><li>Works, but slow — often on the order of ~1 minute vs a few seconds with LM Studio.</li><li>Fine if you refuse a separate LLM server; otherwise prefer LM Studio.</li></ul>MiniMax-oriented skills (I2V / FL2V / R2V) are included so enhanced prompts speak H3’s language (<code>&lt;Picture N&gt;</code>, first/last alignment, reference jobs, soundscape fields).<hr /><h2 id="what's-in-the-box">What’s in the box</h2><ul><li>Lazy Global Selector — single mode control</li><li>Lazy Image Loaders — role + hard-gate by mode</li><li>Lazy Subject &amp; Scene Automation — optional subjects/scenarios/LoRAs</li><li>LazyPrompt — Prompt Engineer — LM Studio / HF / TextGenerate CLIP</li><li>Lazy MiniMax All-in-One — T2V ↔ R2V conditioner auto-switch</li><li>Lazy Model Switcher — fl2va ↔ ref2va</li><li>Native MiniMax H3 decode → video + stereo audio → save</li></ul>Built for people who want one graph, not four copies of the same sampler.<hr /><h2 id="requirements-(short)">Requirements (short)</h2><ul><li>ComfyUI 0.30.0+ (native MiniMax H3)</li><li><a target="_blank" rel="ugc" href="https://github.com/vsaan212/Vsaan212-workflow-utilities">Vsaan212-workflow-utilities</a> (Lazy nodes)</li><li>KJNodes (Set/Get, helpers), plus usual Comfy core loaders/sampler/save</li><li>Models from <a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/MiniMax-H3">Comfy-Org/MiniMax-H3</a>: fl2va + ref2va UNETs, Qwen3-VL encoder, video VAE, audio VAE</li></ul><hr /><h2 id="quick-start">Quick start</h2><ol><li>Install custom nodes + download the H3 files into <code>diffusion_models</code> / <code>text_encoders</code> / <code>vae</code>.</li><li>Open the workflow; set Lazy Global Selector to your mode.</li><li>Drop images on the matching loaders only.</li><li>Write a short idea — enhance with LM Studio (recommended) or bypass for raw text.</li><li>Set duration on MiniMax All-in-One → Queue.</li></ol>

Files with SHA256: 9738d11901a009d8e36a558cc9f40a5f5fcd620f11044b08f2a5c90c344f2286