---
title: SeedVR2-3B-FP8-e4m3fn
canonical_url: "https://www.modelscope.cn/models/theMindStudio/SeedVR2-3B-FP8-e4m3fn"
md_url: "https://www.modelscope.cn/models/theMindStudio/SeedVR2-3B-FP8-e4m3fn.md"
repository: theMindStudio/SeedVR2-3B-FP8-e4m3fn
last_updated: 2026-01-09
license: apache-2.0
base_model:
  - ByteDance-Seed/SeedVR2-3B
base_model_relation: finetune
parameters: 3.6B
tensor_type:
  - F16
  - F8_E4M3
library_name:
  - safetensors
downloads: 2241
stars: 5
---

# SeedVR2-3B-FP8-e4m3fn

> SeedVR2-3B-FP8-e4m3fn - theMindStudio 在 ModelScope 开源的模型。SeedVR2 Optimized Weights (FP8 + FP16 VAE)

theMindStudio/SeedVR2-3B-FP8-e4m3fn 是 ModelScope 魔搭社区上的 3.6B 参数机器学习模型，采用 apache-2.0 许可，基于 ByteDance-Seed/SeedVR2-3B 构建。

- **Repository**: theMindStudio/SeedVR2-3B-FP8-e4m3fn
- **License**: apache-2.0
- **Parameters**: 3.6B
- **Base model**: ByteDance-Seed/SeedVR2-3B
- **Downloads**: 2241
- **Stars**: 5
- **Last updated**: 2026-01-09

Source: https://www.modelscope.cn/models/theMindStudio/SeedVR2-3B-FP8-e4m3fn

---

# SeedVR2 Optimized Weights (FP8 + FP16 VAE)

This repository provides a curated set of weights for the **SeedVR2** upscaler, optimized for high-quality inference with a reduced memory footprint on modern hardware.

## Original Model & Research
- **Title**: SEED-VR: Image-to-Video Upscaling with Specialized Research
- **Original Repo**: [ByteDance-Seed/SeedVR2-3B](https://huggingface.co/ByteDance-Seed/SeedVR2-3B)
- **Organization**: ByteDance (Seed Story Team)

## Included Weights
The files in this repository were adapted from the ComfyUI distribution by [numz/SeedVR2_comfyUI](https://huggingface.co/numz/SeedVR2_comfyUI) to provide a streamlined experience for MindCraft Studio.

- **`seedvr2_ema_3b_fp8_e4m3fn.safetensors`**: The main 3B parameter transformer model quantized to **FP8 (e4m3fn)**. This allows the model to run on consumer hardware with significantly less VRAM while maintaining state-of-the-art upscaling quality.
- **`ema_vae_fp16.safetensors`**: The Exponential Moving Average (EMA) VAE kept in **FP16**. This ensures that the final image generation avoids the artifacts often introduced by low-precision VAEs.

## Usage
These files are intended to be used as a pair within the MindCraft Studio generation engine or any compatible MLX/Python environment.

## License
As with the original upstream model from ByteDance, these weights are distributed under the **Apache License 2.0**.