MiniMax H3 Turbo LoRAs (4–8 Step)
A follow-up to my MiniMax H3 workflow: there are now two turbo LoRAs that cut sampling from ~20 steps down to 4-8 — roughly a 4x speedup — while keeping H3's native audio. This post covers both, with attached workflows for each and a straight recommendation on which to use. If you haven't set up H3 yet, start with the full workflow and memory guide first:
✓ 100% free — download the workflow and run it locally.Requirements
- ComfyUI (portable or desktop), updated to a recent build
- 3 custom node(s), listed below
- NVIDIA GPU recommended; see the notes for model-specific VRAM guidance
Model downloads
- larryvrh/MiniMax-H3-Turbo-Lora — repo https://huggingface.co/larryvrh/MiniMax-H3-Turbo-Lora/tree/main
- Kijai/MiniMax-H3_comfy — loras https://huggingface.co/Kijai/MiniMax-H3_comfy/tree/main/loras
Custom nodes
Setup steps
- Steps & Strength (Applies to Both)
- 4 steps — fast, and genuinely usable on the Larryvrh LoRA. In my testing the Larryvrh version held sharp detail and clean audio even at 4. On the lightx2v LoRA, 4 is more of a fast-draft mode (softer detail).
- 6 steps — the safe balanced default for either LoRA. Full speed benefit, solid detail, clean audio.
- 8 steps — go here if you want the cleanest possible audio. At the very lowest step counts the native audio degrades first, so bump to 8 when audio quality is critical.
Notes & tips
- Which Turbo LoRA Should You Use?: Larryvrh Turbo LoRA — my recommendation. This is the one I use. It's been actively updated and keeps fixing issues from earlier versions, and in my side-by-side testing it produced cleaner, sharper results at fewer steps — Larryvrh at 4 steps looked just as good if not better to me than the lightx2v LoRA at 6 but the lightx2v did sound better. The tradeoff: it needs one extra custom node pack, which changes two nodes in the workflow (the loader and the sampler).; lightx2v LoRA (Kijai conversion) — the simplest setup. Loads natively with a standard LoRA loader, no extra custom nodes. Slightly softer results in my testing, but the easiest to drop in if you want zero workflow changes.
- Option A - Larryvrh Turbo LoRA (Recommended): minimax_h3_turbo_v4_step600.safetensors — my pick (non-EMA). Sharpest results in my testing.; minimax_h3_turbo_v4_step600_ema.safetensors — the EMA variant; smoother and more stable. Worth trying both to see which you prefer.; Several earlier versions are also on the repo if you want to compare.
- Option B - lightx2v Turbo LoRA (Kijai Conversion): minimax_h3_fl2v_lightx2v_turbo_4step_v0.1_comfy.safetensors — full version.; ..resizedavg_rank_21_bf16.safetensors — rank-reduced (300 MB), reported to keep 95% of the quality. Good if you're tight on storage or VRAM.
More resources
- https://www.patreon.com/TheLocalLab/posts/free-minimax-h3-165774435
- https://get.runpod.io/Minimax-H3-ComfyUI
- https://www.youtube.com/@TheLocalLab
- https://www.patreon.com/TheLocalLab
- locallabdigest.com https://locallabdigest.com
- https://www.locallabdigest.com
- Related video — MiniMax H3 video + audio walkthrough https://youtu.be/Idg1PqBcr5c
Want the 1-click version?
Skip the manual model downloads and node installs — get a ready-to-run installer, or unlock all 75+ with Local Lab Pro.