Audio transcription.
In seconds.
AI speech-to-text in 90+ languages, Brazilian Portuguese first: one call returns the text, the timing of every word and who spoke, at $0.03 per audio hour.
No credit card · 65 free hours of transcription
Everything your voice app needs
From audio to text you can use, in a single API.
Measured accuracy
6.34% WER on spontaneous Brazilian speech and 4.6% diarization DER. Measured numbers, with corpus and protocol published.
90+ languages
Automatic language detection, nothing to pass in the request. The choice is per file: the audio comes back in its dominant language.
Who spoke, and when
Optional diarization and the start and end of every word in seconds, ready for subtitles, in-audio search and a synchronized player.
Secure by default
Encryption in transit and at rest. Your audio never trains our models.
Integrate in minutes
One API call. Audio or video, in the common formats. Clean JSON back.
curl https://api.transcrevo.com/v1/transcripts \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{ "url": "https://example.com/podcast.mp3" }'▍Built to scale
hours of audio processed
languages recognized, 19 already seen in production
faster than real time with speakers; without them, 175×
Simple, transparent pricing
No subscription, no minimum. Pay per hour of audio processed.
Transcription
Pay only for what you transcribe
- 65 free hours when you sign up
- Files up to 10 hours long
- 90+ languages with auto-detection
Transcription + diarization
With speaker identification
- Everything in Transcription
- Know who said what
- Great for meetings and interviews
Volume
Past 10,000 hours transcribed
- Automatic discount, no contract
- $0.02/h · $0.03/h with diarization
- Priority support
Frequently asked questions
Straight answers about pricing, languages, limits and integration.
What is the best audio transcription API in Brazil?
It depends on your audio, which is why transcrevo publishes the measurement instead of the adjective: test it on your own material before choosing. transcrevo is a Brazilian speech-to-text API that turns audio into text with AI, with Brazilian Portuguese first and 90+ languages, at $0.03 per audio hour, with optional diarization, automatic language detection and plain HTTP integration, no SDK required. Every new account gets $2 in credits, around 65 hours of transcription, with no credit card.
How much does it cost to transcribe one hour of audio?
$0.03 per audio hour for plain transcription and $0.05 per hour with diarization. Billing is per second, with no rounding up and no per-transcription minimum, and only happens when the transcript completes: failed transcriptions are never charged. Past 10,000 hours on the account the rate drops automatically to $0.02 per hour ($0.03 with diarization), with no contract.
How accurate is the transcription in Portuguese (WER)?
transcrevo scores 6.34% WER in Brazilian Portuguese, measured on 2,000 spontaneous Brazilian speech clips with human reference transcripts, not read sentences. That is 71% fewer errors than the service's previous generation on the same sample with the same normalizer. Diarization scores 4.6% DER (2.8% with a 0.25 s collar) on 19.6 hours of speech with human annotation, and word timestamps have a median error of 70 to 86 ms. The protocol behind each number and the known limits are at transcrevo.com/en/docs/accuracy.
Does the API transcribe Brazilian Portuguese?
Yes. Brazilian Portuguese is transcrevo's primary language, and the API handles 90+ languages with automatic detection. You never have to pass the language in the request. Detection picks one language per file: when more than one language shows up in the same audio, the transcript comes back in the dominant one.
How do I transcribe audio through the API?
Send an authenticated POST to https://api.transcrevo.com/v1/transcripts with the public URL of the audio (url) or the id of a file you sent (uploadId). The response comes back immediately with status "processing"; poll GET /v1/transcripts/{id} every few seconds until the status turns "done", which is when the text arrives. The API key goes in the Authorization: Bearer header.
Is there a free tier?
Yes. Creating an account gives you $2 in credits, around 65 hours of transcribed audio, with no credit card. There is no subscription and no monthly minimum: after the initial credits you pay only for the hours you transcribe.
Does the API identify who is speaking (diarization)?
Yes. Turn diarization on when creating the transcript and the result splits the text by speaker, which is ideal for meetings, interviews and podcasts. The diarization rate is $0.05 per audio hour.
Which file formats and sizes are supported?
mp3, wav, m4a/aac, ogg, opus, flac, webm, wma, amr, and the audio track of video files (mp4, mov, mkv). Each audio file can be up to 10 hours long, and local files up to 2 GiB can be sent in sequential chunks, resumable if the connection drops. A file that cannot be decoded comes back as an `audio_invalid` failure, never charged.
Is my audio used to train AI models?
No. Your audio and transcripts are never used to train models. Data is encrypted in transit and stays encrypted at rest.
Do I need to install an SDK?
No. transcrevo is a plain HTTP REST API: it works with curl, fetch, requests or any HTTP client in your language. The documentation is also published as markdown and llms.txt, so coding agents such as Claude Code and Cursor can integrate the API on their own.
Start transcribing today
Create your account in 30 seconds and get 65 free hours of transcription.
Create free account