AudioNotes
A local AI transcription and notes app uses FunASR and Ollama to turn recordings into Markdown notes and answer questions without cloud AI processing.
Alternative to:
AudioNotes turns audio and video into transcripts and structured Markdown notes on your computer. It's for reviewing meetings, interviews, courses and voice memos while keeping the recordings and resulting notes locally.
- Transcription and notes: Upload audio or video and keep the full transcript alongside organized Markdown notes.
- Recording questions: Ask follow-up questions about the content and return to saved notes and conversation history.
- Browser recording: Capture audio directly in the browser and supply domain-specific vocabulary for speech recognition.
- Language options: The default FunAudioLLM/Fun-ASR-Nano-2512 model prioritizes Chinese recognition; FunAudioLLM/Fun-ASR-MLT-Nano-2512 provides a multilingual alternative.
FunASR handles speech recognition locally, and Ollama runs the model for notes and questions, using qwen3.5:2b by default. Recordings, transcripts, notes and conversation history stay on the computer by default. Downloading the application and models requires internet access; subsequent content processing doesn't depend on cloud AI services. The browser interface accepts access only from the same computer by default.
AudioNotes is open source under the MIT license and runs through Docker or Python. It supports CPU processing, with optional NVIDIA GPU acceleration on Linux. The Docker setup recommends at least 12 GB of memory, or 16 GB for longer recordings, and about 10 GB of disk space. Noise, accents and overlapping speakers can affect transcription accuracy.