Hi FunASR team,
Thank you for the great work on FunASR. It has become one of the most comprehensive speech model collections.
I maintain audio.cpp, a high-performance C++ inference framework for audio models built on GGML. It currently supports 40+ model families across ASR, TTS, voice conversion, music generation, and other audio tasks.
The framework provides shared model components, unified loading and inference APIs, multiple hardware backends, and common performance optimizations that can benefit many models at once. For example, a recently merged framework-level optimization improved the Metal performance of multiple models by up to 60%.
I would love to see the FunASR model family integrated into audio.cpp. Since your team already maintains Llama.cpp- and vLLM-based implementations and is much more familiar with the model architectures and preprocessing pipelines, an implementation from your team would likely be more complete and maintainable than duplicating that work independently.
I would be happy to help with the audio.cpp framework, answer integration questions, and review or contribute to the PR.
Thanks again for the amazing work!
Repo: https://github.com/0xShug0/audio.cpp
Hi FunASR team,
Thank you for the great work on FunASR. It has become one of the most comprehensive speech model collections.
I maintain audio.cpp, a high-performance C++ inference framework for audio models built on GGML. It currently supports 40+ model families across ASR, TTS, voice conversion, music generation, and other audio tasks.
The framework provides shared model components, unified loading and inference APIs, multiple hardware backends, and common performance optimizations that can benefit many models at once. For example, a recently merged framework-level optimization improved the Metal performance of multiple models by up to 60%.
I would love to see the FunASR model family integrated into audio.cpp. Since your team already maintains Llama.cpp- and vLLM-based implementations and is much more familiar with the model architectures and preprocessing pipelines, an implementation from your team would likely be more complete and maintainable than duplicating that work independently.
I would be happy to help with the audio.cpp framework, answer integration questions, and review or contribute to the PR.
Thanks again for the amazing work!
Repo: https://github.com/0xShug0/audio.cpp