Voice Generation

Voice generation APIs turn text or runtime speech content into voice output for voice-enabled flows and telephony integrations.

Key concepts

Term	Definition
Voice generation	Producing an audio response from text or runtime prompt content.
Voice settings	The workflow or request-level settings that select voice, language, and speech behavior.
Voice interaction	A phone or voice-channel conversation that can include generated audio and recordings.
Recording URL	A presigned URL used to retrieve a stored voice interaction recording when available.

What you can do

Action	Endpoint
Generate voice	`POST /llm-chain/voice/generate`
Retrieve a recording URL	`GET /telephony/recordings/:id/presigned-url`

Generate voice

curl -X POST "$OCTO_BASE_URL/llm-chain/voice/generate" \
  -H "Authorization: Bearer $OCTO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "text": "Thanks for calling. I can help with that.",
    "context": { "spaceId": "space_123", "language": "en-US" }
  }'

Voice settings are normally configured on the AI Workflow version under language, voice, and phone settings. Use the API when a trusted integration needs to generate speech outside a normal entry-point call flow.

Retrieve a recording URL

Voice interactions can expose recordings in the Interactions UI when recording is enabled. Integrations can request a presigned URL for a recording by ID.

curl "$OCTO_BASE_URL/telephony/recordings/recording_123/presigned-url" \
  -H "Authorization: Bearer $OCTO_API_KEY"

Treat presigned URLs as temporary sensitive links. Do not store them permanently or expose them outside approved support and compliance workflows.

Operational checks

Situation	What to check
No audio generated	Confirm text input, voice settings, language, and provider availability.
Recording missing	Confirm recording was enabled on the workflow phone settings before the interaction.
Recording URL expired	Request a fresh presigned URL.
Voice call not reaching workflow	Check the phone number, voice entry point, and published workflow version.

FAQ

Should I generate voice directly for every phone call? No. Voice entry points and workflow phone settings handle normal call audio. Use this API for trusted integration paths.

Are recording URLs permanent? No. Treat them as temporary access links and request a new URL when needed.

Phone Numbers - purchase and assign inbound voice numbers
Entry Points - route phone calls to published workflows
Interactions - listen to call recordings and inspect transcripts