ThinkSound

🌐 English | Español | Français | Japanese

       

If you find this project useful,
a star ⭐ on GitHub would be greatly appreciated!

---

Repository layout

This ThinkSound GitHub repository hosts two related projects on separate branches:

Branch	Project	Documentation
master	ThinkSound (NeurIPS 2025) — unified Any2Audio generation with CoT-guided flow matching	This file: README.md
prismaudio	PrismAudio — follow-up work (ICLR 2026) on video-to-audio with multi-dimensional CoT-RL	README.md on the prismaudio branch

For ThinkSound, use branch master (this README). For PrismAudio, check out prismaudio and follow README.md there.

---

ThinkSound is a unified Any2Audio generation framework with flow matching guided by Chain-of-Thought (CoT) reasoning.

PyTorch implementation for multimodal audio generation and editing: generate or edit audio from video, text, and audio, powered by step-by-step reasoning from Multimodal Large Language Models (MLLMs).

📰 News

• 2026.03.24   🔥 PrismAudio is released in the same repo on branch prismaudio — see README.md there for setup and models.
• 2026.01.26   🎉 PrismAudio accepted to ICLR 2026 Main Conference (code/docs on prismaudio).
• 2025.11.25   🔥 Online PrismAudio Demo is live.
• 2025.11.25   🔥 PrismAudio paper on arXiv — multi-dimensional CoT-RL for video-to-audio.
• 2025.09.19   🎉 ThinkSound accepted to the NeurIPS 2025 Main Conference!
• 2025.09.01   Our AudioCoT dataset is now open-sourced and available on Hugging Face!
• 2025.07.17   🧠 Finetuning enabled: training and finetuning code is now publicly available, along with clear usage instructions to help you customize and extend ThinkSound with your own data.
• 2025.07.15   📦 Simplified installation and usability: dependencies on PyPI for easy cross-platform setup; Windows .bat scripts automate environment creation and script running.
• 2025.07.08    🔧 Major update: model lightweighted and optimized memory and GPU usage, now supports high-throughput audio generation at scale!
• 2025.07.01   Online demo on Hugging Face Spaces and ModelScope for interactive experience!
• 2025.07.01   Released inference scripts and web interface;
• 2025.06   ThinkSound paper released on arXiv!
• 2025.06   Online Demo is live - try it now!

---

Follow-up: PrismAudio (same repo, prismaudio branch)

PrismAudio is the successor to ThinkSound (ICLR 2026), developed under a new name but kept in this repository on branch prismaudio. Installation, checkpoints, and citation are in README.md on that branch.

👉 git checkout prismaudio or open the branch on GitHub.

---

🚀 Features

• Any2Audio: Generate audio from arbitrary modalities — video, text, audio, or their combinations.
• Video-to-Audio SOTA: Achieves state-of-the-art results on multiple V2A benchmarks.
• CoT-Driven Reasoning: Chain-of-Thought reasoning for compositional and controllable audio generation via MLLMs.
• Interactive Object-centric Editing: Refine or edit specific sound events by clicking on visual objects or using text instructions.
• Unified Framework: One foundation model supports generation, editing, and interactive workflow.

---

✨ Method Overview

ThinkSound decomposes audio generation and editing into three interactive stages, all guided by MLLM-based Chain-of-Thought (CoT) reasoning:

1. Foley Generation: Generate foundational, semantically and temporally aligned soundscapes from video.
2. Object-Centric Refinement: Refine or add sounds for user-specified objects via clicks or regions in the video.
3. Targeted Audio Editing: Modify generated audio using high-level natural language instructions.

---

⚡ Quick Start

Environment Preparation:

# ThinkSound code: branch master. PrismAudio: clone with -b prismaudio (see README.md on that branch).
git clone -b master https://github.com/liuhuadai/ThinkSound.git
cd ThinkSound
conda create -n thinksound python=3.10
conda activate thinksound
pip install thinksound
conda install -y -c conda-forge 'ffmpeg<7'
# Download pretrained weights https://huggingface.co/liuhuadai/ThinkSound to Directory ckpts/
# model weights can be also downloaded from https://www.modelscope.cn/models/iic/ThinkSound
git lfs install
git clone https://huggingface.co/liuhuadai/ThinkSound ckpts
# To improve inference and training speed, you may optionally install a FlashAttention backend compatible with your system and PyTorch version.

✅ Windows Tip:
Windows users can simply run setup_windows.bat (or double-click it) to automatically create the conda environment, install all dependencies (including FFmpeg), and download the pretrained model — no manual setup required.
Make sure conda and git are installed and available in your system PATH before running the script.

▶️ Run the Demo

Linux/macOS

chmod +x scripts/demo.sh
./scripts/demo.sh <path-to-your-demo-video> <title> <CoT description> [use-half]

Windows

You can use the provided .bat script instead:

.\scripts\demo.bat <path-to-your-demo-video> <title> <CoT description> [use-half]

Note:

• <path-to-your-demo-video>: The path to a single video
• [use-half] (optional): Add use-half at the end to enable half precision feature extraction.

---

📦 Batch Inference

Linux/macOS

chmod +x scripts/eval_batch.sh
./scripts/eval_batch.sh <video_path> <csv_path> <save_path (optional)> [use-half]

Windows

Use the equivalent .bat script:

.\scripts\eval_batch.bat <video_path> <csv_path> <save_path (optional)> [use-half]

Note:

• <video_path>: Path to the root directory containing all .mp4 videos to be processed (all videos must be of equal duration).
• <csv_path>: A CSV file with text prompts for each video (see demo_test.csv for format).
• <save_path> (optional): Where to save generated audio. Defaults to results/features.
• [use-half] (optional): Add use-half at the end to enable half precision feature extraction.

---

Web Interface Usage

For an interactive experience, launch the Gradio web interface:

python app.py

🏋️ Train the Model

See Training.md

---

📄 License

This project is released under the Apache 2.0 License.

Note: The code, models, and dataset are for research and educational purposes only. Commercial use is NOT permitted. For commercial licensing, please contact the authors.

📦 Third-Party Components

• Stable Audio Open VAE (by Stability AI): This repository includes a fine-tuned VAE from Stable Audio Open, licensed under the Stability AI Community License. Commercial use and redistribution require prior permission from Stability AI.

• 📘 All other code and models are released under the Apache License 2.0.

---

Acknowledgements

Many thanks to:

• stable-audio-tools (by Stability AI): For providing an easy-to-use framework for audio generation, as well as the VAE module and weights.
• MMAudio: For the implementation of the MM-DiT backbone in the audio domain.

---

📖 Citation

If you find our project useful in your research or work, please cite our paper:

@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}, 
}
@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}, 
    }

---