Whisper transcription online, with nothing to install
Whisper is an open speech model you can run yourself, if you’re happy to set up Python and wait on your own hardware. Textspert runs open models, Whisper among them, on our own GPUs, and wraps an editor, speaker labels and exports around the result.
Transcribe a recording
Launching November 2026
Which model transcribes your file
Two speech models share the work, and the language decides which one you get.
- Parakeet handles its 25 European languages, English included. We picked it for those languages after testing both models on the same recordings.
- Whisper large-v3-turbo handles all the others, from Japanese and Hindi to Arabic and Swahili.
- Sortformer, a separate model, works out who is speaking. Whisper on its own doesn’t do that.
There’s no model picker, so if you want Whisper’s own output for an English file, this isn’t the tool. We also tried the bigger Whisper large-v3 on our test set. It wasn’t more accurate than turbo and needed several times the GPU time, so turbo stayed.
Running Whisper yourself versus using Textspert
| Question | Whisper on your own machine | Textspert |
|---|---|---|
| Setup | Python, the model files and some patience; a GPU makes it far faster | A browser and an email address |
| Cost | Free software; you pay in hardware and electricity | Free plan, hour packs from $5, or $20 a month |
| Speaker labels | Not built in; needs a second tool | Up to 4 speakers per file, on every plan |
| Editing | A text or subtitle file you fix by hand | An editor where clicking a word plays that moment |
| Exports | Text and subtitle formats | TXT, DOCX, PDF, SRT, VTT, CSV and JSON |
| Privacy | Audio never leaves your computer | Uploaded to our storage and GPU servers; no AI company gets it |
With a capable GPU, a taste for the command line and a need for plain text only, running Whisper locally is a fine choice, and the most private one there is. Textspert is for people who would rather upload a file and get a finished, labeled transcript to work on.
A long file on a laptop CPU can also tie the machine up for a good while, fans roaring. On our side the GPUs do the heavy lifting while you close the tab and come back later. Paid files and hour packs go to the front of the queue; Free files wait their turn when it’s busy.
Transcribing a file with open models, step by step
Create an account
The Free plan needs only an email address. No card, no trial clock.
Upload audio or video
MP3, M4A, WAV, MP4, MOV, MKV, FLAC and other common formats all work. Free files can run up to 60 minutes each.
Let it pick the language
The first 30 seconds of speech set the language, and the file goes to Parakeet or Whisper from there.
Edit and export
Fix names with find and replace, check any word by clicking it, then download the format you need.
Open models on hardware we run
A big reason people look for Whisper is control over where their audio goes. Here, the models run on our own GPUs. Your audio and transcripts are never sent to any AI company, and we never use your files to train or tune a model.
That isn’t the same as running everything on your own laptop, and we won’t pretend it is. The private transcription page explains our privacy promises and deletion. The full list of supported languages has its own page.
Whisper questions
Is Textspert just Whisper with a website?
No. Whisper large-v3-turbo handles languages outside Parakeet’s 25 European languages. Parakeet does the rest, Sortformer labels speakers, and the editor, exports and AI notes are our own work.
Can I choose Whisper for an English file?
No. English goes to Parakeet, the model we chose for English after testing both, and there’s no setting to override that.
Does Textspert use the large-v3 model?
No. We tested it against turbo on our recordings, it wasn’t more accurate, and it took far more GPU time.
Do I need a GPU or Python?
No. Transcription runs on our own GPUs. You only need a browser and the file.
Is it free?
The Free plan covers 3 files a day, each up to 60 minutes, with speaker labels and every export format. See free transcription for the full list.
Related tools
- LanguagesThe 100 transcription languages and how detection works.
- Private transcriptionYour files, sharing and deletion stay in your control.
- Free transcription3 files a day, up to 60 minutes each, with speaker labels and every export format.
- MP3 to textTranscribe MP3, M4A, WAV, FLAC and other audio files without converting them first.
Compare plans on the pricing page, or read our short privacy promises.
Join the waitlist
Launching November 2026. Be first to hear when Textspert opens.
We'll only email you about the launch. Unsubscribe anytime. Privacy