---
license: apache-2.0
pipeline_tag: text-to-video
tags:
- comfyui
- workflow
- minimax-h3
- text-to-video
- video
- audio
- turbo
- lora
---

# MiniMax H3 — ComfyUI Workflows

Drag-and-drop ComfyUI workflows for [MiniMax H3](https://www.minimax.io/blog/minimax-h3), the omni-modal video model that generates video **with native stereo audio** (voice, SFX, room tone — all in one pass). Shared by [@BennyDaBall_OG](https://x.com/BennyDaBall_OG).

## Workflows

| File | What it does |
|---|---|
| [`MiniMax_H3_Turbo_v4_8step_T2VA_BennyDaBall.json`](./MiniMax_H3_Turbo_v4_8step_T2VA_BennyDaBall.json) | Text → video+audio with the **Turbo v4 LoRA** — 8 steps instead of 20, ~**1.9× faster**, audio stays clean |

## Turbo v4 · 8-step T2VA

100% core ComfyUI nodes — **no custom node packs**. Verified with a same-seed A/B against the stock 20-step recipe: same choreography, healthy audio (peak ≈ −9 dB, zero clipping).

**The recipe**

| Setting | Value |
|---|---|
| LoRA | `minimax_h3_turbo_v4_step600_ema_pruned_comfyui` @ **1.0** (built-in *LoraLoaderModelOnly*) |
| Steps / sampler / scheduler | **8 · euler · beta** (no sigma-shift node needed) |
| Guidance | CFG-free (*BasicGuider*) — write a `Negative:` line inside the prompt itself |
| Resolution | 1344×768 @ 24 fps |
| Length | frames on a 17k+5 grid: **124 ≈ 5 s**, 243 ≈ 10 s, 362 ≈ 15 s (model max) |

The bundled demo prompt ships with a **fixed seed** so your first render reproduces the demo clip — flip the seed widget to *randomize* for new takes.

## Getting started

1. Update ComfyUI to **≥ 0.30** (MiniMax H3 nodes are built in).
2. Drag the `.json` onto the ComfyUI canvas.
3. ComfyUI will offer to **download any missing models automatically** (every loader node carries its download URL). Manual links below.
4. Queue. First run: the console should show **zero** `lora key not loaded` warnings.

## Models

| File | Size | Folder |
|---|---|---|
| [`minimax_h3_fl2va_pruned_int8_convrot.safetensors`](https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/diffusion_models/minimax_h3_fl2va_pruned_int8_convrot.safetensors) | 19.5 GB | `models/diffusion_models` |
| [`qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors`](https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/text_encoders/qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors) | 14.6 GB | `models/text_encoders` |
| [`minimax_h3_video_vae_fp16.safetensors`](https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/vae/minimax_h3_video_vae_fp16.safetensors) | 1.4 GB | `models/vae` |
| [`minimax_h3_audio_vae_fp32.safetensors`](https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/vae/minimax_h3_audio_vae_fp32.safetensors) | 0.4 GB | `models/vae` |
| [`minimax_h3_turbo_v4_step600_ema_pruned_comfyui.safetensors`](https://huggingface.co/drbaph/MiniMax-H3-Turbo-Lora-ComfyUI/resolve/main/minimax_h3_turbo_v4_step600_ema_pruned_comfyui.safetensors) | 620 MB | `models/loras` |

## Tested / not tested

- Verified on **5-second clips at 1344×768**. Higher res (1920×1088) and longer clips are untested with this LoRA — earlier turbo versions fell apart there, so treat that as experimental.
- Few-step audio blowout is the classic turbo failure mode; this v4 @ 8 steps is the combo that passed. If you drop steps further, check your audio peaks.

## Prompt pattern that works

Style contract → timed beats (`0.0-0.6s: ...`) → camera lock line → an audio timeline starting with room tone at 0.0s → `Negative:` line. Dialogue inside `<d>[English] ... </d>` is spoken verbatim with lip-sync. Give the first spoken line ≥ 1.2 s of lead-in and anchor t=0 with room tone. The demo prompt in the workflow shows the full pattern.

## Credits

- **MiniMax H3** by [MiniMax](https://www.minimax.io/) · ComfyUI-repackaged weights by [Comfy-Org](https://huggingface.co/Comfy-Org/MiniMax-H3)
- **Turbo LoRA** trained by [larryvrh](https://huggingface.co/larryvrh/MiniMax-H3-Turbo-Lora) (v4_step600 EMA) · ComfyUI convert by [drbaph](https://huggingface.co/drbaph/MiniMax-H3-Turbo-Lora-ComfyUI)
- Workflow + testing by [@BennyDaBall_OG](https://x.com/BennyDaBall_OG)

*License note: this repo covers the workflow files only. The H3 model weights and the LoRA carry their own licenses — check the linked repos before commercial use.*