Local speech to text

Turn audio into clean transcripts with live modes and speaker diarization.

Transcribe (desktop + CLI) uses Whisper to capture audio from files, microphones, or system audio and produce live transcripts with diarization and timestamps. Summarization after capture is optional via Ollama with any Ollama model.

Platforms Windows, macOS, Linux
Modes File, Mic, Speaker
Extras Diarization, Summaries

Real time capture

See speakers and timestamps while recording. Export transcripts to text files automatically.

Offline friendly

Models download on first use and then run locally. Summaries are optional if you run Ollama.

Built for teams

Batch file runs, fast CLI workflows, and consistent output for review or archiving.

Why Transcribe

Everything you need for dependable transcripts.

Capture audio from multiple sources, tag speakers, and deliver export ready text in minutes.

Multi source capture

Record files, microphones, or system output with optional dual source conversations.

Speaker labeling

Add diarization for clearer transcripts and easy handoff to editors or reviewers.

Timestamped output

Every segment includes time markers for quick navigation and review.

Summaries on demand

Pipe results into Ollama to generate summaries after transcription finishes.

Cross platform GUI

Avalonia based desktop app for Windows, macOS, and Linux.

CLI automation

Scriptable commands for batch runs, CI workflows, and power users.

Modes

Pick the capture style that fits your workflow.

File mode

Transcribe a WAV file and export a timestamped transcript.

File based

Mic mode

Live microphone capture with live output and timestamps.

Real time

Speaker mode

Capture system audio plus optional microphone input for conversations.

Dual source

Downloads

Grab the desktop app or CLI.

Latest builds are published as GitHub releases. Packages are self contained and ready to run.

Latest release: open on GitHub.

Desktop GUI

Full interface for file and live capture.

CLI

Scriptable commands for automation and pipelines.

Workflow

From capture to transcript in three steps.

1

Choose a mode

Pick file, microphone, or speaker capture and select your audio device.

2

Transcribe locally

Whisper runs locally after the first model download.

3

Export or summarize

Save timestamped text and optionally generate a summary via Ollama.

System audio tips: macOS may require a loopback driver like BlackHole. Linux uses monitor sources in PulseAudio or PipeWire.

Summaries

Ollama configuration

Summaries are optional. To use them, Ollama must be installed and running locally.

1

Install Ollama

Download and install Ollama for your platform from ollama.com.

2

Run the server

Start it locally with ollama serve.

3

Choose a model

Pull any model with ollama pull <model> before summarizing.

Required: Summaries only work while Ollama is running. The default host is http://localhost:11434, but you can point Transcribe to any Ollama host.

Troubleshooting: confirm ollama serve is running, the model is pulled, and nothing blocks localhost:11434.