HenestrosaDev/audiotext

Audiotext – Desktop transcription, translation & summarisation tool

What it does – Audiotext is a cross‑platform desktop application that lets you turn audio or video into text. It can:

• Import files, YouTube links, or record from the microphone.
• Run Whisper X locally (CPU or CUDA‑accelerated GPU) or call OpenAI’s Whisper, Google Speech‑to‑Text, or other cloud APIs.
• Identify speakers, give word‑level timestamps, and clean out music/background noise.
• Translate the transcript with a range of LLM providers (OpenAI, Claude, Gemini, DeepSeek, Mistral, Grok, Ollama, DeepL, Google Translate).
• Summarise the content into key points/chapters.
• Export to many formats (plain text, Markdown, Word, SRT/VTT subtitles, TSV, JSON).
• Keep a searchable history, tags, notes and a queue for batch processing.
• Offer a full CLI for scripting.

Why it matters – It bundles a powerful speech‑to‑text stack (Whisper X + faster‑whisper + pyannote for diarisation) into a user‑friendly GUI, while keeping data private – everything runs on your machine unless you opt‑in to a cloud API. The built‑in translation and summarisation leverage LLM APIs, turning a raw transcript into a multilingual, bite‑size product without extra tools.

Key technologies

• UI – CustomTkinter (Tkinter‑based) with drag‑and‑drop support.
• ASR engine – WhisperX (uses faster‑whisper, CTranslate2, PyTorch, optional CUDA).
• Speaker diarisation – pyannote.audio.
• Cloud APIs – OpenAI Whisper, Google Speech‑to‑Text, plus LLMs for translation/summarisation (OpenAI, Anthropic Claude, Gemini, DeepSeek, Mistral, Grok, Ollama, DeepL, Google Translate).
• Media handling – FFmpeg + pydub for extracting audio, pytubefix for YouTube downloads, python‑sounddevice for mic I/O.
• Export – python‑docx, built‑in exporters for subtitle formats.
• Packaging – Native installers for macOS, Windows, Linux; includes optional GPU‑accelerated bundles.

Who it’s for – Anyone who needs to transcribe interviews, lectures, podcasts, or video content locally, especially when privacy matters. The multilingual UI and export options make it handy for content creators, researchers, and accessibility work.

Getting started – Download the appropriate installer from the latest GitHub release (macOS DMG, Windows EXE, Linux tar.gz). FFmpeg is bundled; on Linux you may need to install libportaudio2 for mic support. After launch you can drag files onto the window, pick a transcription engine, adjust settings, and hit Transcribe.

Where to learn more – Full docs, tutorials, and language‑specific help are at https://getaudiotext.com. The repo also contains a CLI (cli.py), test suite, and contribution guidelines.