Sending a second text chunk to the same TTS websocket context with continue_=True immediately stopped the ongoing audio generation instead of continuing sequentially. The docs on contexts did not make the pattern obvious. Another user found the fix: pass continue_=True on every context.send() call, including the very first chunk, not just the follow-ups. The reporter confirmed this resolved it.

When the second chunk kills the first chunk's audio, you are probably only setting continue_=True on the follow-up sends. Pass continue_=True on every context.send(), including the first chunk, and the audio generates sequentially across chunks. Check the Cartesia docs page on TTS websocket contexts for the full pattern. The reporter confirmed this fixed it after a user pointed out the first-chunk detail. Source: https://github.com/cartesia-ai/cartesia-python/issues/59