The WhisperAI speech-to-text API turns audio and video into text through one REST endpoint. Send a file up to 5 GB or a public URL, and get JSON back with the transcript, speaker labels and word-level timestamps, in 100+ languages, for $0.01 a minute.
curl https://api.whisperai.com/v1/transcript \
-H "Authorization: wai_your_api_key_here" \
-H "Content-Type: application/json" \
-d '{
"audio_url": "https://example.com/interview.mp4",
"speaker_labels": true,
"webhook_url": "https://your-app.com/hooks/stt"
}'Turn features on with a parameter. Results arrive on the same transcript, so there is no second service to call.
Accurate text from files, URLs or a live stream.
language_codespeaker_labelskeyterms_prompt/v1/realtime/wsStructure on top of the words, in the same job.
sentiment_analysisentity_detectioniab_categoriesauto_highlightsKeep personal and sensitive data out of your app.
redact_piiredact_pii_audiocontent_safetyfilter_profanityThe output format your product needs.
words/{id}/srt/{id}/sentences/{id}/translateSome intelligence features run in a subset of languages. See the API docs for the full parameter list.
Six common products, and the parts of the API each one leans on.
Transcribe hour-long recordings in one request and attribute every line to a speaker.
Generate timed caption files for a video library, then translate them for other markets.
Score calls by sentiment and pull out the names, products and topics customers mention.
Make an archive of episodes searchable down to the word and surface the key moments.
Turn interviews in many languages into speaker-labelled transcripts ready for coding.
Stream microphone audio and show the words on screen while the user talks.
Three steps from an audio file to structured JSON your app can use.
POST /v1/transcript
Pass a public audio_url, or upload a file of up to 5 GB to /v1/upload first. Add the features you want as parameters. The job is queued at once.
webhook_url or GET /v1/transcript/{id}
The status moves from queued to processing to completed. Set a webhook_url to be called back, or poll the transcript.
/sentences /paragraphs /srt /vtt
Read the text, speaker utterances and per-word timestamps from one response, or fetch sentences, paragraphs and subtitle files.
| { | |
| "id": "tr_8fK2…", | |
| "status": "completed", | // when done |
| "language_code": "en", | // auto-detected |
| "audio_duration": 1834, | // seconds |
| "text": "Thanks for joining. Let's start…", | |
| "utterances": [ | // who spoke when |
| { "speaker": "A", "start": 240, "end": 4180, | |
| "text": "Thanks for joining. Let's start…" }, | |
| { "speaker": "B", "start": 4420, "end": 9050, … } | |
| ], | |
| "words": [ | // times in ms |
| { "text": "Thanks", "start": 240, "end": 520, "confidence": 0.98 }, | |
| … | |
| ] | |
| } |
Every field is documented in the speech-to-text API reference.
Eight questions worth asking any provider, with WhisperAI's answers.
Can a full meeting or lecture go through as one request?
Up to 5 GB per file, or pass a public URL. No compressing or splitting.
Does it detect the language and cover your users?
100+ languages with auto-detection for recordings, 18 for realtime, translation into 86.
Is diarization part of the same call, or a separate model?
Same request: set speaker_labels to true.
Can you teach it product names and jargon?
keyterms_prompt takes up to 1,000 words in six major languages, 200 in the rest.
Can you redact personal data and delete results?
PII redaction for text and audio; DELETE /v1/transcript/{id} removes a transcript.
Do you have to poll, or does it call you back?
Either: poll the transcript, or pass webhook_url and get notified.
How many files can run at once?
Up to 250 concurrent requests on both plans.
Per minute, per hour, or a commitment?
$0.01 a minute with no commitment, or $99 a month for 10,000 minutes.
Comparing providers? See WhisperAI vs Deepgram and how it compares with OpenAI's Whisper API.
Start pay as you go or choose Developer for included pre-recorded minutes. Realtime and translation are metered separately on both plans.
| Pay as you go | Developer | |
|---|---|---|
| Pre-Recorded Audio | ||
| Price | $0.01 / minute | $99 / month |
| Included Minutes | - | 10,000 / month |
| Additional Usage | $0.01 / minute | $0.006 / minute |
| REST API | Included | Included |
| Concurrency limit | Up to 250 concurrent requests | Up to 250 concurrent requests |
| Languages | 100+ | 100+ |
| Speaker Labels | Included | Included |
| Timestamps | Included | Included |
| File size limit | 5GB / 10 hours | 5GB / 10 hours |
| Realtime Speech-to-Text | ||
| Price | $0.0067 / minute | $0.0067 / minute |
| Streaming API | Included | Included |
| Realtime Results | Included | Included |
| Languages | 18 | 18 |
| Speaker Labels | Included | Included |
| Translation | ||
| Price | $0.001 / minute / language | $0.0005 / minute / language |
| Availability | Included | Included |
| Target Languages | 86 | 86 |
| Speaker Labels Preserved | Included when speaker labels are enabled on the source transcript | Included when speaker labels are enabled on the source transcript |
| Multiple Languages Per Request | Included | Included |
| Create account | Choose Developer | |
Translation is billed separately from transcription. Speaker labels are preserved in translated utterances when they were enabled on the source transcript.
Protect customer data with encrypted transport and storage, private account access, and revocable API keys.
API traffic is TLS-secured, and stored audio, transcripts, and metadata are encrypted at rest.
Uploaded files remain private to your API account, and transcripts can be deleted through the API.
Create and revoke API keys from the Developer Portal, and keep secret keys on your server.
Pricing, languages, limits and privacy.
Create an account, get a key in the Developer Portal, and send your first file in minutes.