text to speech
Generating spoken audio from text with a voice model.
In more detail
Providers limit input length per request and offer fixed voices; OpenAI's policy requires telling listeners the voice is AI-generated. Check the output with an independent transcriber, since a voice can skip or mangle words.
Related terms
- speech to text — Converting spoken audio into text, often with a timestamp for each word.
- speech to speech — A voice pipeline in which one model listens and answers in audio directly, as opposed to chaining speech to text, a text model and text to speech.
- EU AI Act — The European Union's law regulating AI systems by risk, with duties for providers and deployers that phase in from 2025.
Taught in
- Cost, Latency And Caching — AI Engineering: LLM Apps, RAG & Agents
- Images, Audio And Speech — AI Engineering: LLM Apps, RAG & Agents
References
- OpenAI: Text to speech (disclosure of AI-generated voices)
- Deepgram: Text to speech REST API (2,000 character input limit for Aura-2)