---
title: PrismAudio
canonical_url: "https://www.modelscope.ai/models/iic/PrismAudio"
md_url: "https://www.modelscope.ai/models/iic/PrismAudio.md"
repository: iic/PrismAudio
last_updated: 2026-03-24
license: mit
pipeline_tag: audio-generation
tasks:
  - audio-generation
base_model:
  - google/videoprism-large-f8r288
  - google/t5gemma-l-l-ul2-it
base_model_relation: finetune
library_name:
  - pytorch
frameworks:
  - PyTorch
language:
  - zh
downloads: 176
stars: 3
tags:
  - audio
  - generation
  - video2audio
  - music
---

# PrismAudio

> PrismAudio - An open-source model by iic on ModelScope.        

iic/PrismAudio is a audio-generation model on ModelScope. licensed under mit. derived from google/videoprism-large-f8r288、google/t5gemma-l-l-ul2-it.

- **Repository**: iic/PrismAudio
- **License**: mit
- **Tasks**: audio-generation
- **Base model**: google/videoprism-large-f8r288, google/t5gemma-l-l-ul2-it
- **Tags**: audio, generation, video2audio, music
- **Downloads**: 176
- **Stars**: 3
- **Last updated**: 2026-03-24

Source: https://www.modelscope.ai/models/iic/PrismAudio

---

<h1 align="center">PrismAudio</h1>
<p align="center">
  <img src="https://img.shields.io/badge/ICLR 2026-Main Conference-blue.svg" alt="ICLR 2026"/>
</p>

<p align="center">
  <a href="https://arxiv.org/abs/2511.18833">
    <img src="https://img.shields.io/badge/arXiv-2511.18833-b31b1b.svg" alt="arXiv"/>
  </a>
  &nbsp;
  <a href="http://prismaudio-project.github.io/">
    <img src="https://img.shields.io/badge/Online%20Demo-🌐-blue" alt="Online Demo"/>
  </a>
  &nbsp;
  <a href="https://github.com/FunAudioLLM/ThinkSound/tree/prismaudio">
    <img src="https://img.shields.io/badge/GitHub-Code-black?logo=github" alt="GitHub"/>
  </a>
  &nbsp;
  <a href="https://huggingface.co/spaces/FunAudioLLM/PrismAudio">
    <img src="https://img.shields.io/badge/HuggingFace-Spaces-orange?logo=huggingface" alt="Hugging Face"/>
  </a>
  &nbsp;
  <a href="https://www.modelscope.cn/studios/iic/PrismAudio">
    <img src="https://img.shields.io/badge/ModelScope-在线体验-green" alt="ModelScope"/>
  </a>
</p>

---

**PrismAudio** 是首个将强化学习融入视频转音频（V2A）生成的框架，并配备专门的思维链（CoT）规划机制。在 ThinkSound 开创性的基于 CoT 的 V2A 框架基础上，PrismAudio 进一步将单一推理分解为四个专门的 CoT 模块（语义、时序、美学和空间），每个模块配备针对性的奖励函数，实现多维强化学习优化，从而在所有感知维度上同步提升推理能力。

---

## 快速开始

详细的训练和推理代码请参考 [ThinkSound（prismaudio 分支）](https://github.com/FunAudioLLM/ThinkSound/tree/prismaudio)。

```bash
git clone -b prismaudio https://github.com/liuhuadai/ThinkSound.git
cd ThinkSound

conda create -n prismaudio python=3.10
conda activate prismaudio
chmod +x scripts/PrismAudio/setup/build_env.sh
./scripts/PrismAudio/setup/build_env.sh

# 下载预训练权重到 ckpts/ 目录
# Hugging Face：https://huggingface.co/FunAudioLLM/PrismAudio
# ModelScope：  https://www.modelscope.cn/models/iic/PrismAudio
git lfs install
git clone https://huggingface.co/FunAudioLLM/PrismAudio ckpts
```

---

## 许可证

本项目基于 [MIT 协议](https://opensource.org/licenses/MIT) 发布。

> **注意：** 代码、模型权重和数据集**仅供研究和教育用途**。未经作者明确授权，不允许商业使用。

---


## 📖 引用

如果 PrismAudio 对您的研究有帮助，请引用我们的论文：

```bibtex

@misc{liu2025prismaudiodecomposedchainofthoughtsmultidimensional,
        title={PrismAudio: Decomposed Chain-of-Thoughts and Multi-dimensional Rewards for Video-to-Audio Generation}, 
        author={Huadai Liu and Kaicheng Luo and Wen Wang and Qian Chen and Peiwen Sun and Rongjie Huang and Xiangang Li and Jieping Ye and Wei Xue},
        year={2025},
        eprint={2511.18833},
        archivePrefix={arXiv},
        primaryClass={cs.SD},
        url={https://arxiv.org/abs/2511.18833}, 
}

@misc{liu2025thinksoundchainofthoughtreasoningmultimodal,
      title={ThinkSound: Chain-of-Thought Reasoning in Multimodal Large Language Models for Audio Generation and Editing}, 
      author={Huadai Liu and Jialei Wang and Kaicheng Luo and Wen Wang and Qian Chen and Zhou Zhao and Wei Xue},
      year={2025},
      eprint={2506.21448},
      archivePrefix={arXiv},
      primaryClass={eess.AS},
      url={https://arxiv.org/abs/2506.21448}, 
}
```

---

## 📬 联系我们

✨ 如有任何问题或建议，欢迎[提交 Issue](https://github.com/liuhuadai/ThinkSound/issues)