Getting started

Pascal API

Speech and audio models in one API.

The base URL is https://api.pascalai.co. Create an API key in the dashboard and send it on every call as Authorization: Bearer. Keys begin with sk_.

A coding agent should read https://www.pascalai.co/llms.txt and connect to the read-only docs server at https://www.pascalai.co/.well-known/mcp. That server searches these docs. It cannot create a key or spend credits. The signed-in connector is https://api.pascalai.co/mcp. It uses a key from this dashboard, shows the price before a transcribe or a speak, and cannot create a key either.

What you can do

Speech to text

Transcribe one audio file with POST /v1/speech-to-text. Submit up to 50 files with POST /v1/speech-to-text/batch.

Speech to text realtime

Transcribe live audio over a WebSocket at GET /v1/speech-to-text/realtime.

Text to speech

Synthesize speech from text with POST /v1/text-to-speech. Submit up to 50 texts with POST /v1/text-to-speech/batch.

Text to speech realtime

Synthesize speech live over a WebSocket at GET /v1/text-to-speech/realtime. The model query parameter selects the stream.

Voice cloning

Speak new text in a voice taken from one reference clip with POST /v1/voice-cloning. Submit up to 50 texts with POST /v1/voice-cloning/batch.

Voice design

Describe a voice, then speak text in that voice, with POST /v1/voice-design. The description is required. Submit up to 50 texts with POST /v1/voice-design/batch.

Voice conversion

Keep the words of a recording and replace the voice with POST /v1/voice-conversion. Send a source clip and a reference clip. Submit up to 50 source clips with one reference clip using POST /v1/voice-conversion/batch.

Voice continuation

Continue a recording with POST /v1/voice-continuation. Send a prompt clip, the transcript of that clip, and the new words.

Stem separation

Split a mix into separate sources. The model id selects the separation.

Voice isolator

Reduce noise and interference in a recording.

Audio enhancement

Reconstruct damaged speech.

Get started

Create a key, then transcribe a file. The walkthrough is on Quickstart.

curl https://api.pascalai.co/v1/speech-to-text \
  -H "Authorization: Bearer $PASCAL_API_KEY" \
  -H "Idempotency-Key: $(uuidgen)" \
  -F model=whisper-large-v3-turbo \
  -F params='{"language":"en"}' \
  -F file=@meeting.wav

curl https://api.pascalai.co/v1/speech-to-text/$JOB_ID \
  -H "Authorization: Bearer $PASCAL_API_KEY"

Create the key in the dashboard under API keys.

Billing

A new account starts with 1,000 credits in the studio. The API balance starts at zero. Keys spend API credit only. API credit is added from the dashboard, from $5, or from ₹499 in India. A platform fee is added on top of the credit: 5.5% of the credit, or $0.80, whichever is larger. In India, GST is added on the credit plus that fee. The fee and the tax do not add credits. API credit stays until it is used.

A successful request is charged a $0.05 fee, which is 50 credits, plus usage. One credit is $0.001. Speech to text, stem separation, voice isolation, and audio enhancement are billed on the duration of the input audio, in minutes. Pascal Transcribe is priced per audio hour. Voice conversion is billed on the duration of the source clip, in minutes. Text-to-speech, voice cloning, voice design, and voice continuation are billed on the duration of the output audio, in minutes, except where the model page prices characters. Those prices are per million characters. A successful speech-to-text file is charged $0.05 plus the audio duration.

Send many files with POST /v1/speech-to-text/batch. Up to 50 files go in one request. A successful batch is charged $0.05 once. Each file is charged for its audio duration in minutes.

Send many texts with POST /v1/text-to-speech/batch, POST /v1/voice-design/batch, or POST /v1/voice-cloning/batch. Up to 50 texts go in one request. A voice-cloning batch sends one reference clip with those texts. A successful batch is charged $0.05 once. Each text is charged for the generated audio duration in minutes. Send many source clips with one reference clip using POST /v1/voice-conversion/batch. Up to 50 source clips go in one request. Each clip must be under 25 MB. A successful batch is charged $0.05 once. Each source clip is charged for its duration in minutes.

Live text to speech at GET /v1/text-to-speech/realtime is charged $0.05 plus the audio produced, at $0.00333 per minute. A session runs for up to 5 hours of audio.

Price a duration before submitting it with GET /v1/models/{id}/quote?seconds=. seconds is the audio duration to price.

Errors

A failed call returns JSON with error.type, error.code, error.message, and error.request_id. Branch on error.code. The codes are listed on Errors.