AI SDK 7.0.150 adds an experimental streaming-speech interface
Released October 22, AI SDK 7.0.150 adds experimental_streamSpeech, converting strings or incoming text parts into speech while exposing audio, full-event, usage, warning and provider-metadata streams.
Why it matters
Speech extends generative interfaces from visual components into live multimodal output, but cancellation, errors and consumption semantics directly affect playback state, caption synchronization and recovery.
A closer look
Input can be a string or an array, iterable or stream of text deltas and text parts, with optional voice and providerOptions per part. The returned audioStream and fullStream are single-consumer, with usage, warnings, responses and providerMetadata promises. Provider error parts are non-fatal and routed to onError, while user abort or cancellation rejects result promises immediately. The release also exports experimental types, adds a cause to NoSpeechGeneratedError and ships a mock streaming-speech model for tests.
Who this is for
AI SDK teams building live spoken responses, cancellable audio playback and multimodal agent interfaces
Original sources
Revisions & corrections (1)
— Official release confirms version, timestamp, API contract and failure semantics; published as a distinct multimodal interface event with experimental and observability limits.