---
language:
- en
- zh
tags:
- human-animation
- talking-head
- video-generation
- nf4
- quantized
- echomimic
base_model: BadToBest/EchoMimicV3
library_name: diffusers
license: apache-2.0
---

# EchoMimicV3 — NF4 Quantized

This is a **NF4 (NormalFloat4) quantized** version of [EchoMimicV3](https://huggingface.co/BadToBest/EchoMimicV3).

## Quantization Details

| Parameter | Value |
|-----------|-------|
| Quantization Type | NF4 (NormalFloat4) |
| Double Quantization | Yes |
| Compute Dtype | float16 |
| Library | bitsandbytes |
| VRAM Reduction | ~65% (5 GB → ~1.7 GB) |

## Usage

```python
from transformers import BitsAndBytesConfig
import torch

nf4_config = BitsAndBytesConfig(
    load_in_4bit=True,
    bnb_4bit_quant_type="nf4",
    bnb_4bit_use_double_quant=True,
    bnb_4bit_compute_dtype=torch.float16
)
```

## Citation

```bibtex
@misc{meng2025echomimicv3,
  title={EchoMimicV3: 1.3B Parameters are All You Need for Unified Multi-Modal and Multi-Task Human Animation},
  author={Rang Meng, Yan Wang, Weipeng Wu, Ruobing Zheng, Yuming Li, Chenguang Ma},
  year={2025},
  eprint={2507.03905},
  archivePrefix={arXiv}
}
```