Gemini Realtime: Transcribe model audio via gemini api #1446
Conversation
🦋 Changeset detectedLatest commit: a6eac32 The changes in this PR will be included in the next version bump. This PR includes changesets to release 1 package
Not sure what this means? Click here to learn what changesets are. Click here if you're a maintainer who wants to add another changeset to this PR |
| @@ -382,7 +382,7 @@ def _on_input_speech_done(self, content: TranscriptionContent) -> None: | |||
| # TODO: implement sync mechanism to make sure the transcribed user speech is inside the chat_ctx and always before the generated agent speech | |||
|
|
|||
| def _on_agent_speech_done(self, content: TranscriptionContent) -> None: | |||
There was a problem hiding this comment.
when interrupted, are we only transcribing until the moment of interruption?
There was a problem hiding this comment.
No, the current implementation transcribes the entire text (all the frames which are received before interruption). It's hard to determine the exact point of interruption since we receive frames faster than the actual playback.
There was a problem hiding this comment.
got it. I think that's fine.. in the v1 branch, the synchronization/truncation logic will be downstream from the model.. model should just produce the entire thing.
Uh oh!
There was an error while loading. Please reload this page.