Does the audio leave my computer?
Not when you use the on-device engine. Transcription runs locally; your recordings are not uploaded to transcribe them.
Speech to text
Turn speech into text without uploading it
Import a recording or a video, and EdgeSpeak transcribes it on your computer while you review it against playback.
Direct answer
EdgeSpeak is a desktop app that transcribes audio and video locally with Lattice-2 Flash or Pro, which support more than 40 languages. Files stay on your machine.
Review while you listen
The transcript follows playback. Click a line to jump to it, fix a word in place, and export when it reads right.
Timestamps and structure
Get word-level timestamps, sentence segmentation and optional speaker labels. Leave the language empty and it is detected from the audio.
Export or automate
Export plain text, SRT, VTT or JSON. The same engine runs from the command line and the OpenAI-compatible local gateway.
edgespeak-cli transcribe meeting.m4a -o meeting.json
Works on video too
Drop in an MP4 or another video file and get the spoken text with timing, ready to become subtitles.
FAQ
Which languages are supported?
Lattice-2 Flash and Pro support more than 40 languages. The language guide lists the codes and what each capability supports.
Can I get timestamps for every word?
Yes. Word-level timestamps are available in the JSON export, the CLI and the local gateway.
Can I call it from my own scripts?
Yes. Use edgespeak-cli, or send audio to the OpenAI-compatible local gateway on your computer.