Introduction
TTS-1 HD renders supplied copy as a narrated recording. OpenAI positions this Speech API model around audio quality, making it a fit for listening material that will be prepared and reviewed before playback. It accepts text rather than a live microphone stream, and the written script determines what is said. The published shutdown date is January 6, 2027.
Key features & highlights
The HD variant provides a fixed set of nine voices for narration: alloy, ash, coral, echo, fable, onyx, nova, sage and shimmer. Voice selection changes the sound of the reader; it does not create a custom speaker identity. English is the optimization target, so pronunciation and pacing in another language should be reviewed on representative passages.
TTS-1 HD supports both delivery-ready compressed recordings and uncompressed audio for an editing workflow. Select MP3, Opus or AAC for a small listening asset, or FLAC, WAV or PCM when the downstream pipeline needs another representation. The speed control can adapt the pace of a prepared passage without changing the character-based billing quantity.
Request parameters
| Parameter | Required | Description |
|---|---|---|
text | Yes | Submit the final TTS-1 HD narration script; a request can contain at most 4,096 characters. at least 1 characters · at most 4096 characters |
voice_id | No | Choose the HD narrator from nine built-in presets; use a sample passage to review names and emphasis. |
speed | No | Adjust TTS-1 HD pace from 0.25 to 4 before generating a section; this does not change its billed character count. |
output_format | No | Select the HD recording format for your publishing pipeline; raw PCM needs explicit 24 kHz, 16-bit little-endian playback settings. |
Pricing
$0.006 / 1,000 characters
Use cases
A learning product can use TTS-1 HD to prepare chapter introductions and explanations that a reviewer listens to before publication. Retaining the approved script beside the recording makes later corrections and accessibility review easier.
For an audio edition of a product guide, generate TTS-1 HD narration by section and combine it in the publishing workflow. Section boundaries provide natural places to insert pauses, replace a revised paragraph or offer the listener a navigation point.
Prompt guide & pro tips
Prepare the HD script as finished editorial copy, with punctuation that expresses the intended phrasing. Resolve ambiguous symbols, uncommon names and long strings of digits before synthesis. Test a short representative passage with the chosen voice and listening speed; a clean transcript alone does not establish that every term is pronounced as intended. Extra style instructions are not a delivery-control feature of this legacy Speech model.
Send at most 4,096 characters in one TTS-1 HD request. For long chapters, divide the script at sentence or section boundaries and assemble the returned files in an audio editor. Keep generated files below 32 MiB, paying attention to uncompressed formats at very slow speeds. Store the script and section order with the approved audio, disclose synthetic narration to listeners, and plan migration using those source materials before the announced retirement.
API quick start
# Generate the idempotency key once; reuse it when retrying after a lost response.
export SPICY_IDEMPOTENCY_KEY="${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: $SPICY_IDEMPOTENCY_KEY" \
-d '{
"model": "openai/tts-1-hd/text-to-speech",
"input": {
"text": "Welcome back. Your files finished rendering a minute ago and are ready to download.",
"voice_id": "alloy",
"output_format": "mp3"
}
}'
# Copy data.taskId from the response into TASK_ID, then poll the status.
TASK_ID="TASK_ID_FROM_RESPONSE"
curl "https://api.spicyapi.ai/api/v1/jobs/recordInfo?taskId=$TASK_ID" \
-H "Authorization: Bearer $SPICY_API_KEY"
See the API reference for authentication, the complete schema, polling, and error handling.

