Introduction
TTS-1 converts a written message into a spoken audio file. Its original design favors quick speech generation for short interactions, such as a status announcement or an application instruction. This is the Speech API model: it reads supplied words, without listening to a caller or composing an answer. OpenAI has scheduled its retirement for January 6, 2027.
Key features & highlights
A TTS-1 request can use alloy, ash, coral, echo, fable, onyx, nova, sage or shimmer. These preset voices are optimized for English while supporting text in other languages. Keep a chosen voice consistent across related messages, and review important names with a listener who knows the intended language.
TTS-1 can produce MP3, Opus, AAC, FLAC, WAV or raw PCM. Compressed output is useful for a spoken notification that must travel over a mobile connection; an uncompressed file fits an application that already controls its audio playback pipeline. Speaking speed is adjustable from 0.25 to 4, but changing speed does not remove words from the submitted message.
Request parameters
| Parameter | Required | Description |
|---|---|---|
text | Yes | Give TTS-1 the exact spoken message, up to 4,096 characters including spaces and punctuation. at least 1 characters · at most 4096 characters |
voice_id | No | Pick a TTS-1 preset from alloy, ash, coral, echo, fable, onyx, nova, sage and shimmer; the default is alloy. |
speed | No | Set TTS-1 speaking speed between 0.25 and 4; 1 keeps the normal pace. |
output_format | No | Choose a TTS-1 file encoding; PCM contains raw 24 kHz, 16-bit little-endian samples without a WAV header. |
Pricing
$0.003 / 1,000 characters
Use cases
For a task dashboard using TTS-1, turn a completed action into a short audible status message. Have the application decide the status and write the exact words before requesting audio, so the announcement cannot invent an outcome.
For a device setup flow, prepare one TTS-1 clip for each instruction or confirmation. Separate clips let the interface play only the relevant step and replace a changed instruction without regenerating the rest of the sequence.
Prompt guide & pro tips
Write TTS-1 input as the sentence a listener should hear. Expand a product abbreviation when pronunciation matters, and test account codes, dates and decimal numbers in context. A direction such as ‘use a calm voice’ is spoken content here; this legacy model does not offer instruction-based delivery control. Use the voice and speed fields for the available controls.
Keep each TTS-1 text within 4,096 characters and each generated file within the platform’s 32 MiB limit. Split long narration into separate messages, especially with slow WAV or PCM output. Check the returned recording before placing it in a user flow, and identify the voice as generated audio in the listening experience. Because the model has an announced retirement date, keep the source text and playback configuration available for migration.
API quick start
# Generate the idempotency key once; reuse it when retrying after a lost response.
export SPICY_IDEMPOTENCY_KEY="${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: $SPICY_IDEMPOTENCY_KEY" \
-d '{
"model": "openai/tts-1/text-to-speech",
"input": {
"text": "Welcome back. Your files finished rendering a minute ago and are ready to download.",
"voice_id": "alloy",
"output_format": "mp3"
}
}'
# Copy data.taskId from the response into TASK_ID, then poll the status.
TASK_ID="TASK_ID_FROM_RESPONSE"
curl "https://api.spicyapi.ai/api/v1/jobs/recordInfo?taskId=$TASK_ID" \
-H "Authorization: Bearer $SPICY_API_KEY"
See the API reference for authentication, the complete schema, polling, and error handling.

