The short answer
Open Transcript Studio, pick the spoken language (or keep “Detect automatically”), choose “Accurate” on a computer and select your audio or video file. When the text appears, read it once, fix the odd word and export it as text, Word document or subtitles.
Open the free toolSee the steps in action
A short recording of our tool, without sound. Turn on captions for the steps in your language. The written instructions below cover the same process.
Open or download the video · 0.2 MB · 24.58 s
Each step is also explained in writing.
- Choose the spoken language, or let the tool detect it.
- On a computer, choose “Accurate” for the cleanest text.
- Choose the example recording. The speech model runs in this browser.
- The words appear as text with time stamps.
- Copy the text or download it as a document or subtitles.
Which recordings work
Transcript Studio opens audio and video files: MP3, M4A, WAV, OGG, OPUS, FLAC, WebM, MP4, MOV and MKV. That covers phone recordings, Zoom and Teams downloads, podcast episodes and voice messages. You can also select “Record with the microphone” and dictate directly.
The clearer the recording, the better the text. One person speaking at a time, little background noise and a phone held close to the speaker make the biggest difference – more than any setting.
Get the text in four steps
Choose the language
Under “Spoken language”, choose the language of the recording. “Detect automatically” works well for one language; choosing it yourself avoids surprises with short clips.
Pick the model
“Fast” downloads about 80–210 MB of model files and suits phones. On a computer, “Accurate” (about 250–590 MB) copes better with accents and noise. The files download once and are reused.
Add the recording
Select “Choose a file” and pick your recording. The speech model runs on your device; with a recent graphics chip an hour of audio takes a few minutes.
Check and export
Read the text with its time stamps, correct names and technical terms, then choose “Copy text”, “Text (.txt)”, “Word (.docx)” or “Subtitles (.srt)”.
Transcribe WhatsApp voice messages
Voice messages are ordinary sound files once they are saved. In WhatsApp, long-press the voice message and use the share option to save it to your phone’s files; on a computer, download it from WhatsApp Web or Desktop. Then choose that file in Transcript Studio. In Chrome and Edge, WhatsApp’s .opus files open directly.
A voice message contains someone else’s words. Only transcribe messages you received yourself, and keep the text as private as the recording.
Choose the right export
For subtitles, “Characters per subtitle line” sets how many characters fit on one line, so the captions stay readable on a phone.
| For | Export |
|---|---|
| Notes, quotes, an email | Copy text or Text (.txt) |
| An interview with time references | Text with times |
| Editing and sharing as a document | Word (.docx) |
| Subtitles for a video | Subtitles (.srt) or (.vtt) |
| Further processing in your own tools | JSON |
Privacy and accuracy
Your recording is processed on your device and never uploaded; only the speech model is downloaded once from Hugging Face. That is why the tool can handle confidential interviews, medical notes or business calls.
No speech model is perfect. Names, rare words and people talking over each other are the usual mistakes. Search the text for important names and listen to the passages that matter before you quote them.
Questions you might still have
Is the transcription really free?
Yes. Transcript Studio is free, without account, minute limits or watermark. Your device does the work, so there is nothing to pay for.
How accurate is it?
With a clear recording, most sentences come out right. Accents, noise and overlapping voices cause more mistakes; the “Accurate” model handles them better than “Fast”.
Can I transcribe a video?
Yes. Choose an MP4, MOV, MKV or WebM file; the tool uses its sound. For subtitles burned into the video, continue in Subtitle Studio.
Sources & useful details
The screen recordings show LoschaZ tools using sample files created for these guides. They are demonstrations, not tests on every phone or promises of an identical result. Examples are not customer files.
