Deepdub is an AI dubbing and localisation company that also exposes its speech technology to developers as a text-to-speech API. The API generates speech from text with controls over tempo, variance, and duration, clones voices from short audio samples, and blends accents between locales. Audio is returned over HTTP or streamed over a WebSocket, which is the mode aimed at voice agents.
The agent-facing offering is built on the Phantom X model family and advertises time-to-first-audio under 250 ms, support for more than 100 languages and dialects, and accent control across 130 or more languages. A REST endpoint and a WebSocket endpoint are available, with official Python and JavaScript/Node.js SDKs that include streaming support. Deepdub's broader platform covers media dubbing, live dubbing, and an orchestration layer for voice, WhatsApp, SMS, and email agents.
Features
- Text-to-speech: generate speech from text with tempo, variance, duration, seed, and pitch controls
- Streaming delivery: WebSocket streaming for real-time agents alongside HTTP responses, with time-to-first-audio under 250 ms
- Voice cloning: replicate a voice from a short sample supplied as an upload or a base64-encoded reference
- Accent blending: mix accents between locales with an adjustable ratio, with accent control across 130+ languages
- Emotion controls: adjust emotional intensity and expressivity for context-aware agent speech
- Voice library: licensed preset voices selectable by locale
- Speech-to-speech: convert and translate an existing voice recording into another voice
- SDKs: Python and JavaScript/Node.js clients with streaming support
- Models: the dd-etts-3.0 model on the REST API and the Phantom X family for agent speech
