Make.com TTS: swap ElevenLabs for Audra HTTP — Audra
Make.com (formerly Integromat) users usually arrive here from an ElevenLabs module they are replacing. The good news: Make's HTTP > Make a request module is strictly more capable than a vendor-specific module — you gain retries, error handlers, and raw response handling. The migration is ten minutes of module rewiring.
Wiring the HTTP module
- Add module: HTTP > Make a request. Method: POST. URL: https://audratalks.com/v2/speech.
- Body type: Raw, Content type: JSON (application/json). Paste the JSON body below and map your script variable into text.
- Headers: Authorization = Bearer sk-audra-xxxx, Content-Type = application/json. Under 'Advanced settings', set 'Request timeout' to at least 180000 ms — Audra renders take 1–3 minutes for typical scripts and the Make default of 40 seconds will abort mid-render.
- Response handling: check 'Parse response' off and set 'Response format' to Text; add an error handler route for 4xx/5xx so one bad module does not halt the scenario.
{
"model": "audra-essential",
"voice": "june",
"text": "{{10.script}}",
"format": "mp3"
}
From the ElevenLabs module: what maps to what
The ElevenLabs Make module exposed a text box, a voice ID dropdown, and returned a binary audio bundle. In the HTTP module, text becomes the text field, the voice ID becomes a short slug like june or wren (browse /voices), and the binary comes from Make's built-in conversion: follow the HTTP module with a 'Tools > Set a variable' is unnecessary — instead add a second module that consumes the response body directly, for example Google Drive > Upload a File, and map 'Data' from the HTTP module. Set the file name to {{voice}}-{{format}}-{{now}}.mp3.
Error handlers and retries, the Make way
Make's advantages over Zapier show up here. Add an error handler route on the HTTP module: a Resume directive for 429 with a 65-second Sleep before it, a Break directive (retry in 15 min) for 5xx, and a Rollback for 402 so the scenario stops before burning operations. Set the scenario's 'Number of consecutive errors' to 5 and 'Max number of cycles' high enough for your batch. For long scripts — anything past 8,000 characters, which the sync endpoint may auto-queue — submit to POST /v2/speech/jobs with webhook_url pointed at a Make custom webhook; Make will receive the completion callback natively, which the ElevenLabs module never offered.
Voice-by-voice comparison before you commit: /voices has previews. The full HTTP field reference is at /docs.