Transcribe in any of 90+ languages, then process across languages — translation, summaries, terminology, and localization prep with free prompts.
Here's what you get — speaker labels, timestamps, and multiple download formats. Try it with your own file.
Multilingual audio is normal — international teams switch languages mid-meeting, research interviews happen in the participant’s language, and global content needs subtitles in several. The workflow that handles all of it: transcribe in the source language first, then process with language-aware prompts, keeping the original transcript as the reference layer rather than translating audio blind.
The prompt library covers the recurring jobs. Translate Transcript to English produces a faithful English working copy while flagging idioms that resist literal translation. Multilingual Meeting Summary summarizes a mixed-language discussion into one language, noting which language each decision was made in. Code-Switching Analysis maps where and why speakers switch languages — research gold for bilingual-communication studies. Cross-Language Terminology keeps technical terms consistent across a multilingual transcript set, and Subtitle Localization Prep structures a transcript for translators with timing, context notes, and on-screen-text flags.
Every prompt is free in our GitHub library, in Markdown for copy-paste and YAML for automation. Transcription itself supports 90+ languages with automatic language detection at the same $2 per audio hour — upload in whatever language the recording happened, and the processing layer takes it from there.
Upload audio in the language it was spoken — detection and transcription handle the rest, speaker labels included.
Translation, mixed-language summaries, code-switching analysis, terminology consistency, and localization prep — free, tested, copy-paste.
Transcribe first, translate second. The original-language transcript stays as your verifiable reference for every downstream output.
Yes — transcription handles the dominant language well, and mixed segments transcribe in the language spoken. The Multilingual Meeting Summary prompt then produces a single-language summary noting the switches.
The transcript. Transcribing in the source language and translating the text preserves a verifiable original — translating audio directly leaves you nothing to check the translation against.
Yes — the full library is open on GitHub in Markdown and YAML, usable with ChatGPT, Claude, Gemini, or any LLM. Transcription is the only paid step, at $2 per audio hour.
Major languages (Spanish, French, German, Mandarin, Japanese, and dozens more) transcribe at accuracy comparable to English. Lower-resource languages vary — the $2 test on a real file answers it for your language pair.
$2 per hour. No subscription. Files are auto-deleted after processing.