Novita AI adds Bland BTTS with free voice cloning and $0.04 per 1,000 characters
Novita AI has made Bland BTTS, the text-to-speech engine from Bland AI, available through a native audio API. According to Novita, synthesis costs $0.04 per 1,000 characters, voice cloning is free, and both endpoints return synchronously.
- Bland BTTS is live on Novita AI as a passthrough proxy to Bland's speech engine, billed through a Novita account.
- Synthesis costs $0.04 per 1,000 characters, Han characters count as two, and voice cloning is free.
- Requests are synchronous, return binary audio and are capped at 5,000 characters.
- Expressiveness and stability controls work only on BTTS V3 voices.
- Voices are referenced by UUID, and cloned voices are shared across all callers using the same API key.
A proxy to Bland's speech engine
Bland BTTS runs on Novita AI as a passthrough proxy, not a natively hosted model. Novita forwards requests to Bland's production speech engine and bills them through the customer's Novita account. The voices and the request schema are Bland's, while authentication uses a single Novita AI API key.
Bland AI is known for voice agents that handle phone calls, and the engine reflects that focus. Novita says BTTS has two selectable generations, V2 and V3. Expressiveness and stability controls apply only to V3 voices. Audio comes back as binary data rather than a JSON job, and the output supports telephony use.
Novita contrasts this with hosted audio models such as Fish Audio S2 Pro, which it serves directly. For teams that specifically want Bland's voice stack, the company says the proxy saves a second vendor integration.
Two endpoints, free cloning, per-character billing
The API exposes two endpoints under the base URL https://api.novita.ai:
- Speech synthesis: POST /v3/bland-tts/v2/tts accepts text and a voice UUID and returns audio bytes.
- Voice cloning: POST /v3/bland-tts/v1/voices/clone accepts a short audio sample and returns a JSON voice_id.
Neither endpoint takes a model field, since routing is determined by the path. Only synthesis is billed: $0.04 per 1,000 characters under the SKU BLAND_BTTS, and Han characters count as two. Cloning generates no charge. Novita notes that some voice APIs charge a one-time fee per clone, and says free cloning means a catalog of branded voices costs only the seconds those voices speak. Pricing was checked on October 23, 2026, and Novita advises checking its pricing page before budgeting production volume.
Limits developers should know
Requests are synchronous, so there is no task ID to poll. Each synthesis request is capped at 5,000 characters of text. Novita says this covers most agent turns but makes the API a poor fit for long-form narration.
Bland BTTS uses its own request schema rather than an OpenAI-compatible one, so an existing Novita text integration needs a separate client path. Voices are referenced by UUID, not display name. Cloned voices are shared among all callers using the same API key, and the API does not record which caller created which voice. Teams sharing a key therefore need their own ownership tracking, and the shared library does not isolate multiple tenants.
Who it is for
Novita positions Bland BTTS for phone and voice agents, IVR prompts and products with many branded voices. It also suits products already on Novita AI that want Bland's engine without a separate Bland integration, and turn-by-turn synthesis where a synchronous response is simpler than a job queue.
For long-form narration, a single OpenAI-style SDK for text and speech, or multilingual cloning with emotion tags, Novita points to hosted alternatives such as Fish Audio S2 Pro as the closer fit.
