Practical overview
When it fits
Use Hayamimi when you want local, CPU-based live transcription and can install Python dependencies and pretrained models. Choose the microphone, WAV, WebSocket, or supported system-audio input that matches the audio source.
Who it is for
- People who need live captions from microphone audio without using a cloud speech API.
- Meeting participants who want transcripts from a microphone and, on Windows, PC audio output.
- Streamers who want to display captions through an OBS browser-source overlay.
- Developers seeking a local transcription pipeline with a browser dashboard and structured runtime controls.
Documented uses
- Display partial captions as speech is happening and finalized transcript lines afterward.
- Transcribe WAV audio through the same recognition pipeline.
- Combine microphone and PC audio output for meeting transcription on Windows.
- Add live speaker labels or translate Japanese lines into supported target languages.
- Present live captions in a browser dashboard or OBS overlay.
Requirements & limits
- Python 3.10 or later and ffmpeg on PATH are required.
- The standard pretrained-model download is about 3.1 GB; the documented Japanese-and-English minimal option is about 1.1 GB.
- Recognition uses quantized ONNX models through sherpa-onnx and is designed for CPU use.
- Windows 11 is the documented development and test environment; macOS and Linux are expected to work but are not yet tested end to end in CI.
- PC audio loopback is documented for Windows only.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Set up Hayamimi for microphone transcription
Install the documented prerequisites and models, then start live transcription from the microphone.
Before you start
- Python 3.10 or later
- ffmpeg available on PATH
- About 3.1 GB for the standard pretrained-model download, or about 1.1 GB for the Japanese-and-English minimal set
Steps
Create a Python virtual environment.
Install the requirements from requirements.txt using the environment's pip.
Run scripts/download_models.py, adding --minimal for the smaller Japanese-and-English model set.
Run scripts/realtime_transcribe.py from the virtual environment to transcribe microphone audio.
Expected outcome
Hayamimi starts real-time speech transcription from the microphone.
Guide limitations
- The README documents separate Windows and macOS/Linux command forms.
- The project says macOS and Linux are expected to work but are not yet tested end to end in CI.
Start the live dashboard and OBS caption overlay
Run Hayamimi with its local browser dashboard and overlay server.
Before you start
- Hayamimi installed with its pretrained models downloaded
- Python 3.10 or later and ffmpeg on PATH
Steps
Start scripts/realtime_transcribe.py with the --serve option from the virtual environment.
Open http://localhost:8833/dashboard in a browser to view the live dashboard.
For stream captions, add http://localhost:8833/ as an OBS Browser Source.
Expected outcome
The local server exposes a live transcription dashboard and an OBS browser-source overlay.
Guide limitations
- The dashboard and overlay depend on the transcription application running with --serve.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 15 October 2026
- Original website title
- oboroge0/hayamimi · Official homepage title · retained in its source language