Pika H3 Flex ComfyUI Workflow | 8-Step Ref2Video
Load VAE
Already loads minimax_h3_audio_vae_fp32.safetensors
ComfyUI model file · VAE (audio)
MiniMax H3 audio VAE (fp32) · Audio VAE for MiniMax H3 video with sound
This is the audio VAE that MiniMax H3 uses to turn its audio latents into a soundtrack. Every H3 workflow that outputs sound needs it next to the video VAE. Put it in ComfyUI/models/vae and load it with Load VAE.
There is only one precision of this file. It stays fp32 because the audio VAE is small and loses quality when cast down.
minimax_h3_audio_vae_fp32.safetensors
The short version
MiniMax H3 generates video and audio together. Its sampler works in two latent spaces, one for frames and one for sound, so ComfyUI needs a separate VAE to decode each. minimax_h3_audio_vae_fp32.safetensors is the audio one.
It is the same file in every H3 variant: T2VA, FL2VA, Ref2VA, pruned or full, Turbo or not. Download it once and reuse it across all H3 workflows.
Install
ComfyUI/ └─ models/ ├─ … └─ vae/ └─ minimax_h3_audio_vae_fp32.safetensors
$ cd ComfyUI/models/vae$ wget -O minimax_h3_audio_vae_fp32.safetensors \$ https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/vae/minimax_h3_audio_vae_fp32.safetensors$ sha256sum minimax_h3_audio_vae_fp32.safetensors# 8e505d95dd1561d47abd43d4238fd40d9bb1ae9e147ed0a4cba778d76ae4db48
Safetensors files hold tensors only and cannot run code. Still compare the SHA256: a failed download often leaves an HTML error page saved under the model’s filename.
Use
H3 workflows load two VAEs side by side: the video VAE for frames and this one for sound. Load VAE reads the file, and its VAE output goes to the audio decode step of the H3 graph.
output●
Two VAEs, two nodes
Keep one Load VAE for minimax_h3_video_vae_fp16 and a second one for this file.
Silent output?
If the video has no sound, check that this VAE is wired to the audio decode node.
Works with
minimax_h3_audio_vae_fp32.safetensors is one part of the stack. A workflow that reports a missing model usually needs these too. Use the Ref2VA diffusion model instead of FL2VA for reference-to-video workflows.
| File | Role | Folder | Size | Download |
|---|---|---|---|---|
| minimax_h3_fl2va_pruned_int8_convrot.safetensors | Diffusion model (FL2VA) | models/diffusion_models/ | 20.97 GB | Get |
| qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors | Text encoder (Qwen3-VL 32B) | models/text_encoders/ | 15.69 GB | Get |
| minimax_h3_video_vae_fp16.safetensors | Video VAE | models/vae/ | 5.21 GB | Get |
| minimax_h3_fl2v_turbo_4step_v1.0_768p_comfyui_bf16.safetensors | Turbo LoRA, 4 steps(optional) | models/loras/ | 1.96 GB | Get |
Run it
We match workflows by the model files they load: this exact file first, then other versions of it, then workflows built for the same use, then other models from the same family.
35 workflows
These workflows already load minimax_h3_audio_vae_fp32.safetensors. Open one and run it as is.
Load VAE
Already loads minimax_h3_audio_vae_fp32.safetensors
Load VAE
Already loads minimax_h3_audio_vae_fp32.safetensors
Load VAE
Already loads minimax_h3_audio_vae_fp32.safetensors
Load VAE
Already loads minimax_h3_audio_vae_fp32.safetensors
Load VAE
Already loads minimax_h3_audio_vae_fp32.safetensors
Load VAE
Already loads minimax_h3_audio_vae_fp32.safetensors
Load VAE
Already loads minimax_h3_audio_vae_fp32.safetensors
Load VAE
Already loads minimax_h3_audio_vae_fp32.safetensors
Load VAE
Already loads minimax_h3_audio_vae_fp32.safetensors
Load VAE
Already loads minimax_h3_audio_vae_fp32.safetensors
Load VAE
Already loads minimax_h3_audio_vae_fp32.safetensors
Load VAE
Already loads minimax_h3_audio_vae_fp32.safetensors
Load VAE
Already loads minimax_h3_audio_vae_fp32.safetensors
Load VAE
Already loads minimax_h3_audio_vae_fp32.safetensors
Load VAE
Already loads minimax_h3_audio_vae_fp32.safetensors
Load VAE
Already loads minimax_h3_audio_vae_fp32.safetensors
Load VAE
Already loads minimax_h3_audio_vae_fp32.safetensors
Load VAE
Already loads minimax_h3_audio_vae_fp32.safetensors
Load VAE
Already loads minimax_h3_audio_vae_fp32.safetensors
Load VAE
Already loads minimax_h3_audio_vae_fp32.safetensors
Load VAE
Already loads minimax_h3_audio_vae_fp32.safetensors
Load VAE
Already loads minimax_h3_audio_vae_fp32.safetensors
Load VAE
Already loads minimax_h3_audio_vae_fp32.safetensors
Load VAE
Already loads minimax_h3_audio_vae_fp32.safetensors
Load VAE
Already loads minimax_h3_audio_vae_fp32.safetensors
Load VAE
Already loads minimax_h3_audio_vae_fp32.safetensors
Load VAE
Already loads minimax_h3_audio_vae_fp32.safetensors
Load VAE
Already loads minimax_h3_audio_vae_fp32.safetensors
Load VAE
Already loads minimax_h3_audio_vae_fp32.safetensors
Load VAE
Already loads minimax_h3_audio_vae_fp32.safetensors
Load VAE
Already loads minimax_h3_audio_vae_fp32.safetensors
Load VAE
Already loads minimax_h3_audio_vae_fp32.safetensors
Load VAE
Already loads minimax_h3_audio_vae_fp32.safetensors
Load VAE
Already loads minimax_h3_audio_vae_fp32.safetensors
Load VAE
Already loads minimax_h3_audio_vae_fp32.safetensors
Compare
| File | Relation | Size | Best for |
|---|---|---|---|
| minimax_h3_audio_vae_fp32.safetensors | This page · audio VAE | 605 MB | Every H3 workflow with sound |
| minimax_h3_video_vae_fp16.safetensors | Companion · video VAE | 5.21 GB | Decoding the frames |
| minimax_h3_video_vae_int8_convrot.safetensors | Companion · smaller video VAE | 2.81 GB | Lower VRAM |
Troubleshoot
Why: The workflow asks for this exact filename, but ComfyUI cannot find it in models/vae. Usually the file is missing, sits in a different folder.
Fix: Move the file into ComfyUI/models/vae/, rename it to match the workflow (or pick it again in the dropdown), then refresh the node list or restart ComfyUI.
Why: The download is incomplete, or the browser saved an HTML sign-in or error page under the model’s filename.
Fix: Check that the file is exactly 605,254,808 bytes and that its SHA256 matches the value above. Download it again from Hugging Face · Comfy-Org/MiniMax-H3 if not.
Rather skip the setup?
Run Pika H3 Flex ComfyUI Workflow in your browser on a cloud GPU. Nothing to download.
FAQ
ComfyUI/models/vae/. Then refresh the node list or restart ComfyUI so Load VAE lists the file.Sources: Comfy-Org/MiniMax-H3 on Hugging Face · MiniMaxAI/MiniMax-H3 (original weights) · ComfyUI MiniMax H3 tutorial
RunComfy is the premier ComfyUI platform, offering ComfyUI online environment and services, along with ComfyUI workflows featuring stunning visuals. RunComfy also provides AI Models, enabling artists to harness the latest AI tools to create incredible art.