Text to Speech API: one endpoint, 100 voices, instant MP3

Most TTS REST APIs make you learn a request envelope, a voice taxonomy, and an audio-encoding matrix before you hear a single word. InstantVoiceAI collapses that into a single call. You POST JSON to one URL, authenticate with one header, and receive the finished MP3 in the response body — synchronously. There is no job queue, no webhook, and nothing to download separately.

Create a key at /api-keys in the dashboard, drop it into the curl command below, and you have a working MP3. Keys start with ivai_ and are shown exactly once at creation time, then stored hashed — copy it into your secrets manager or environment when you create it. Here is a complete, copy-pasteable request:

curl -X POST https://instantvoiceai.com/api/v1/tts -H "Authorization: Bearer ivai_YOURKEY" -H "Content-Type: application/json" -d '{"text":"Hello world","voice":"azure-aria"}' --output hello.mp3

That writes a playable hello.mp3 to your working directory. Swap the voice id, change the text, and you are integrated. Note that API access requires a Pro or Studio plan.

The request body takes three fields. `text` is the string to speak. `voice` is any id from the catalog (for example azure-aria). `speed` accepts normal, slow, or fast. The response is the audio itself: a binary MP3 stream with Content-Type audio/mpeg, ready to write to a file, return to a browser, or push to object storage. Because the response is plain MP3 bytes, it slots into any HTTP client in any language — no custom parsing required.

The API is not a stripped-down subset. Every voice and language available in the InstantVoiceAI web app is reachable through the same /api/v1/tts endpoint, powered by Microsoft Azure and Google neural models. Browse the full set on the voices page, copy the id you want, and pass it as the voice field. That means you can localize an app across 29 languages without touching multiple providers or stitching together different voice vendors.

API calls draw from your account's monthly character allowance — the same pool the web app uses — so cost is a function of characters, not opaque credits or per-request surcharges. Pro gives you 2,000,000 characters a month (plus 200,000 premium HD-voice characters) and Studio gives you 4,000,000. There is no separate API meter to reconcile, no premium markup for using the endpoint instead of the UI, and a one-time 100,000-character top-up ($8, never expires) is there if a busy month runs long. For high-volume voiceover pipelines, that flat model is typically far cheaper per character than usage-billed competitors.

Anything you have configured in your pronunciation dictionary applies to API output too. If you have taught the app to say a brand name, an acronym, or a tricky proper noun a specific way, the same custom replacements are applied to the text before synthesis on every API call. You define the rules once in the dashboard and they govern both the UI and the endpoint — so generated audio stays consistent no matter which path created it.

A synchronous MP3 endpoint with flat character pricing fits a wide range of automated audio work. Because there is no job queue or callback to manage, it is especially well suited to scripts, build steps, and request-time generation where you just want bytes back.

InstantVoiceAI is not trying to be the most configurable TTS platform in existence — it is trying to be the one you can integrate before lunch and budget without a spreadsheet. Compared with the ElevenLabs API, Amazon Polly, and Google Cloud Text-to-Speech, you trade a sprawling options surface for one endpoint, one auth header, and a flat monthly character allowance instead of per-request or credit-based metering. For teams that mainly need good neural voices, many languages, and predictable cost, that simplicity is the feature.