Some text is replaced by later parts of the sentence

[logiota]

Some text is replaced by later parts of the sentence,
generally where there should be a comma.

I said: please let me know where I can find green apples
And after different text appearing the only text left was :
green apples

great work btw!

[danmaevsky]

I am getting the same issue right now. It doesn't appear to be updating like it does in the demo GIF. It is replacing the transcription line with a text snippet rather than replacing it entirely with the new phrase.

[danmaevsky]

I figured it out!
We need to keep track of the phrase and concatenate the audio stream to it until we detect that we are in a new phrase. Here is how to do it:

Before the while loop, add this line:
audio_data = b''

In the while loop, add the same line to the conditional that checks if a phrase has been completed, so that now it looks like this:

# If enough time has passed between recordings, consider the phrase complete.
# Clear the current working audio buffer to start over with the new data.
if phrase_time and now - phrase_time > timedelta(seconds=phrase_timeout):
phrase_complete = True
audio_data = b''

Lastly, replace the line that says
audio_data = b''.join(data_queue.queue)
with
audio_data = audio_data + b''.join(data_queue.queue)

In effect, this now allows the model to make a new inference on the old data + continuation of the data instead of just the continuation chunk of the data.