Introduction
GPT-4o Transcribe Diarize turns a recorded conversation into text with speaker labels and time ranges. The result lets a reviewer locate the audio for a particular utterance and see when the floor changed hands. Choose the file task for an existing recording where the order of speakers matters to the reader.
Key features & highlights
Each returned segment has a start, an end and a speaker label, so a player can jump to a particular part of the conversation without estimating time from the number of words. A label distinguishes voices within this recording. It does not establish anyone's identity or link voices across recordings.
Automatic chunking splits longer audio at speech boundaries. The finished task keeps segment order alongside the continuous transcript, allowing an application to render a readable conversation and still seek back to the exact part represented by a label. Review crosstalk and quick interruptions before attributing quotations to named participants.
Request parameters
| Parameter | Required | Description |
|---|---|---|
audio_url | Yes | Conversation to split into speaker turns, up to 25 MB at least 1 characters |
language | No | Two-letter language of the recording |
temperature | No | Sampling temperature between 0 and 1 |
Pricing
$0.0001 / second of input audio
Use cases
A podcast editor can search the conversation for a passage, jump to the associated start time and listen to the exchange before marking a clip for editing. A producer can keep the speaker label until a participant confirms attribution.
In a usability interview, group participant responses separately from the moderator's questions. Review transitions where both people spoke at once before linking a quotation to a finding.
Prompt guide & pro tips
Record a short exchange in which two voices alternate before uploading a full interview. Read the completed segments in sequence, compare their time ranges to the original file and listen to the first transition. If a speaker appears to change halfway through an utterance, keep the original audio available for a human review.
Submit a direct recording or a platform upload within the file size and duration limits. The service applies automatic segmentation, including for files longer than thirty seconds. The customer quote is based on recording length, so the number of speaker labels, segments or words does not increase the price of that recording.
API quick start
# Generate the idempotency key once; reuse it when retrying after a lost response.
export SPICY_IDEMPOTENCY_KEY="${SPICY_IDEMPOTENCY_KEY:-$(uuidgen)}"
curl -X POST https://api.spicyapi.ai/api/v1/jobs/createTask \
-H "Authorization: Bearer $SPICY_API_KEY" \
-H "Content-Type: application/json" \
-H "Idempotency-Key: $SPICY_IDEMPOTENCY_KEY" \
-d '{
"model": "openai/gpt-4o-transcribe-diarize/speech-to-text",
"input": {
"audio_url": "https://example.com/reference.mp4"
}
}'
# Copy data.taskId from the response into TASK_ID, then poll the status.
TASK_ID="TASK_ID_FROM_RESPONSE"
curl "https://api.spicyapi.ai/api/v1/jobs/recordInfo?taskId=$TASK_ID" \
-H "Authorization: Bearer $SPICY_API_KEY"
See the API reference for authentication, the complete schema, polling, and error handling.




















