LiveKit Community

What is the difference between synthesizestream and chunkstream class?

Thanks for reaching out, Aditya. I'm an AI assistant and I'm here to help. Give me a moment while I research this topic.


The difference between the *SynthesizeStream* and *ChunkedStream* classes lies in their purpose and usage:
• *ChunkedStream*: This is used with the `synthesize()` method when you have a complete text to convert to audio at once. It requires `input_text` at initialization and is not tracked in a collection of streams.
• *SynthesizeStream*: This is used with the `stream()` method for ongoing or real-time text-to-speech conversion. It does not require `input_text` at initialization, is added to a collection of streams, and includes a `_segments_ch` attribute for word streams.
Both classes handle text-to-speech conversion but are designed for different use cases in the API.
 Sources: <https://github.com/livekit/agents/blob/06a1c61a250769814cdab99114aadb69fbbbd4be/livekit-plugins/livekit-plugins-playai/livekit/plugins/playai/tts.py#L124-L182|GitHub::livekit::agents | livekit-plugins/livekit-plugins-playai/livekit/pl...>

does that mean it directly accepts on streaming input tokens and generates audio in realtime

Hey there! So, the *SynthesizeStream* class is pretty cool - it can actually take streaming text and turn it into audio on the fly. Imagine typing something and hearing it spoken almost instantly! It works with Resemble AI's WebSocket API, which basically means it can break down your text into sentences or words and generate audio in real-time. Super neat, right?
 Sources: <https://github.com/livekit/livekit_composite/blob/4b437f6db613b41c9f1fb00a45d3f11b4a24970b/livekit/agents/livekit-plugins/livekit-plugins-resemble/livekit/plugins/resemble/tts.py#L222-L271|GitHub::livekit::livekit_composite | livekit/agents/livekit-plugins/livekit...>