100% Local · 100% Private
Speech-to-Text,
Your Way
Longform & live transcription, speaker diarization, AI summaries, and a calendar-based audio notebook — powered by Whisper, NeMo, SenseVoice, VibeVoice & more. Runs entirely on your machine.
Hardware Acceleration
Features
Everything You Need
Transcribe
Record anything. Keep everything local.
Record for as long as you want — from your microphone or system audio — and get the full transcript seconds after you stop, with a rolling preview while you speak. Flip on Live Mode for real-time, sentence-by-sentence dictation, or translate foreign audio to English on the fly.
30 minutes of audio in under a minute on an RTX 3060.
Your Hardware
Any GPU. Your choice of models.
Whisper, NVIDIA NeMo Parakeet & Canary, SenseVoice, VibeVoice-ASR, whisper.cpp, MLX — pick what fits your machine. The app manages the whole engine for you: Docker on Linux and Windows, native Metal on Apple Silicon, with guided setup checks and one-click image updates.
Understand
Who said what — automatically.
Speaker diarization labels every voice in the recording. Generate an AI summary of any note, or chat about it with the AI Assistant — wired to any OpenAI-compatible provider, from local LM Studio and Ollama to Groq and OpenRouter.
Organize
A calendar for your voice.
Every recording can land in the Audio Notebook: browse by month, scrub through your day hour by hour, replay the original audio, and find any spoken phrase again with full-text search.
And more
- Remote access — Tailscale or LAN
- Global shortcuts — dictate into any app
- File import — audio & video → .txt / .srt / .ass
- OpenAI-compatible API — drop-in STT for other tools
- Outgoing webhooks — push transcripts anywhere
- Translation — 90+ languages to English
See It In Action
Watch the Demo
Three and a half minutes from launch to diarized transcript.
Backstory
About This Project
TranscriptionSuite started as a personal tool and turned into a hobby project. I’m an engineer ‐ just not a software engineer. This whole thing is vibecoded, but not blindly: for example, Dockerizing the server for easy distribution was 100% my idea.
I’m using this project to learn programming. Starting from virtually nothing, I now have a decent grasp of Python, git, uv & Docker. I started doing this because it’s fun, not to make money ‐ though I do find, despite my mechanical engineering degree, that I want to follow software as a career.
Since I dogfood the app every day, I’m not going to abandon it. I’ll also try to deal with bugs as soon as possible.
Inspired by RealtimeSTT .
Open Source & Free
This project was only made possible thanks to free & open source software.
I wanted to share back to the community and that's why I chose the GPLv3+ license.
Star the repo, report bugs, or contribute on GitHub.
