# h3_upscaler_lms_v0.1.safetensors

> H3 Upscaler LMS v0.1 · Sharpness-tuned latent upscaler for MiniMax H3

This is an experimental fine-tune of the MiniMax H3 3D latent upscaler by Alissonerdx, trained for 2,000 steps on the same sharpness data as his LMS LoRA. Put it in ComfyUI/models/latent_upscale_models and select it in the Minimax H3 Latent Upscaler (3D) node.

This is the recommended fp16 file. The _fp32 file in the same folder is the training artifact, twice the size, and not needed for normal use.

## Facts

- **Type:** Latent upscale model (fine-tune)
- **Format:** Safetensors · fp16
- **Size:** 691 MB (690,593,048 bytes)
- **SHA256:** 40d4228227146245ef2b65480b647c50222465ca125de5880f192e3259e40d59
- **ComfyUI folder:** ComfyUI/models/latent_upscale_models/
- **Loader node:** Minimax H3 Latent Upscaler (3D)
- **License:** Apache-2.0 (base upscaler)

## What is h3_upscaler_lms_v0.1.safetensors?

Alissonerdx started from LBH-123-AI’s minimax_h3_latent_upscaler_3d_fp16 and fine-tuned it on a sharpness dataset, the one behind his minimax_h3_lms LoRA. The goal is latent upscaling that also adds crisp detail.

He calls it a research checkpoint: it can improve perceived detail, but it does not rebuild text, logos or fine structure that are already missing. Because the architecture is unchanged, it loads in the same 3D upscaler node as the original.

## Download and install

- [Hugging Face · Alissonerdx/Minimax-H3-ComfyUI](https://huggingface.co/Alissonerdx/Minimax-H3-ComfyUI/resolve/main/latent_upscaler/h3_upscaler_lms_v0.1.safetensors) · saves as `h3_upscaler_lms_v0.1.safetensors` · Author’s repo. Recommended fp16 file.

```sh
cd ComfyUI/models/latent_upscale_models
wget -O h3_upscaler_lms_v0.1.safetensors \
  https://huggingface.co/Alissonerdx/Minimax-H3-ComfyUI/resolve/main/latent_upscaler/h3_upscaler_lms_v0.1.safetensors
sha256sum h3_upscaler_lms_v0.1.safetensors
# 40d4228227146245ef2b65480b647c50222465ca125de5880f192e3259e40d59
```

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.

## How to load it in ComfyUI

The upscaler works on latents, between a low-resolution H3 pass and a refining pass at the target size. The Minimax H3 Latent Upscaler (3D) node loads the file and upscales in one step.

- **Why latent upscaling:** It skips the decode → pixel upscale → encode round trip through H3’s large video VAE.
- **Scales:** The 3D model was trained at 2×, 2.25×, 2.5× and 4×; its author reports weaker results below 1.5×.

## Files a MiniMax H3 video workflow needs

Use the Ref2VA diffusion model instead of FL2VA for reference-to-video workflows.

- [`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) · Diffusion model (FL2VA) · `models/diffusion_models/` · 20.97 GB
- [`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) · Text encoder (Qwen3-VL 32B) · `models/text_encoders/` · 15.69 GB
- [`minimax_h3_video_vae_fp16.safetensors`](https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/vae/minimax_h3_video_vae_fp16.safetensors) · Video VAE · `models/vae/` · 5.21 GB
- [`minimax_h3_audio_vae_fp32.safetensors`](https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/vae/minimax_h3_audio_vae_fp32.safetensors) · Audio VAE · `models/vae/` · 605 MB
- [`minimax_h3_fl2v_turbo_4step_v1.0_768p_comfyui_bf16.safetensors`](https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/loras/minimax_h3_fl2v_turbo_4step_v1.0_768p_comfyui_bf16.safetensors) · Turbo LoRA, 4 steps · `models/loras/` · 1.96 GB

## Workflows: Uses another version of this file

These workflows load a different precision or format of the same model. Select h3_upscaler_lms_v0.1.safetensors in their Minimax H3 Latent Upscaler (3D) node to use this version.

- [MiniMax H3 Action Scenes | Guided Action Video](https://www.runcomfy.com/comfyui-workflows/minimax-h3-action-scenes-reference-video-native-audio)

## Workflows: Uses other MiniMax H3 models

These workflows load another MiniMax H3 model. Select h3_upscaler_lms_v0.1.safetensors in their Minimax H3 Latent Upscaler (3D) node to use it instead.

- [MiniMax H3 Singularity Reference To Video ComfyUI Workflow](https://www.runcomfy.com/comfyui-workflows/minimax-h3-singularity-reference-to-video-comfyui)
- [MiniMax H3 Latent Upscaler | BUNNYH3 Two-Pass Workflow](https://www.runcomfy.com/comfyui-workflows/minimax-h3-latent-upscaler-comfyui-two-pass-3d-video)
- [MiniMax H3 8-Step Acceleration | All-in-One Video Workflow](https://www.runcomfy.com/comfyui-workflows/minimax-h3-8-step-acceleration-all-in-one-comfyui-workflow)

## Same file, other names, and alternatives

- `h3_upscaler_lms_v0.1.safetensors` · This page · fp16 · 691 MB · Normal inference
- `h3_upscaler_lms_v0.1_fp32.safetensors` · Same model · fp32 training artifact · 1.38 GB · Archival; not needed in ComfyUI
- `minimax_h3_latent_upscaler_3d_fp16.safetensors` · Base model · LBH-123-AI · 691 MB · Faithful upscaling without extra sharpening

## Common errors and fixes

### Value not in list: model_name: 'h3_upscaler_lms_v0.1.safetensors' not in [...]

Why: The workflow asks for this exact filename, but ComfyUI cannot find it in models/latent_upscale_models. Usually the file is missing, sits in a different folder.

Fix: Move the file into ComfyUI/models/latent_upscale_models/, rename it to match the workflow (or pick it again in the dropdown), then refresh the node list or restart ComfyUI.

### Missing node type: MinimaxH3LatentUpscaler3D

Why: Minimax H3 Latent Upscaler (3D) is not built into ComfyUI. It comes from the Comfyui_Minimax_h3_latent_Upscaler custom node pack.

Fix: Install Comfyui_Minimax_h3_latent_Upscaler with ComfyUI Manager (or git clone it into ComfyUI/custom_nodes), then restart ComfyUI.

### SafetensorError: Error while deserializing header

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 690,593,048 bytes and that its SHA256 matches the value above. Download it again from Hugging Face · Alissonerdx/Minimax-H3-ComfyUI if not.

## FAQ

### fp16 or fp32?

fp16. The author recommends it for the included workflows and normal inference; fp32 is kept only as the full-precision training file.

### Where do I put h3_upscaler_lms_v0.1.safetensors in ComfyUI?

In `ComfyUI/models/latent_upscale_models/`. Then refresh the node list or restart ComfyUI so Minimax H3 Latent Upscaler (3D) lists the file.

### Which ComfyUI node loads h3_upscaler_lms_v0.1.safetensors?

**Minimax H3 Latent Upscaler (3D)**. It comes from the [Comfyui_Minimax_h3_latent_Upscaler](https://github.com/LBH-123-AI/Comfyui_Minimax_h3_latent_Upscaler) custom node pack, so install that first.

### Can I use h3_upscaler_lms_v0.1.safetensors commercially?

The repo’s custom LoRA licenses do not cover this upscaler; it keeps the terms of the model it was fine-tuned from, LBH-123-AI’s Apache-2.0 latent upscaler. The MiniMax H3 base model keeps its own license.

## Sources

- [Alissonerdx/Minimax-H3-ComfyUI](https://huggingface.co/Alissonerdx/Minimax-H3-ComfyUI)
- [Latent upscaler notes](https://huggingface.co/Alissonerdx/Minimax-H3-ComfyUI/blob/main/docs/latent-upscaler.md)

Canonical page: https://www.runcomfy.com/comfyui-workflows/model-files/h3_upscaler_lms_v0.1.safetensors
