diff options
| author | yum <yum.food.vr@gmail.com> | 2023-06-26 00:58:58 -0700 |
|---|---|---|
| committer | yum <yum.food.vr@gmail.com> | 2023-06-26 01:46:42 -0700 |
| commit | 0ed379f2c99ac5c126a6f101965ef1eaa58c017b (patch) | |
| tree | 057ea9d58991ab71cd4f6eb4dfb33693f8914e03 /Scripts | |
| parent | e35abb6b9a04c6bebd6875901dbf0671f5cd860d (diff) | |
Add browser source, hardcoded to port 8097
Transcription output now streams to localhost:8097.
In OBS:
* Create a browser source.
* url: localhost:8097
* width: 2200
* height: 400
TODO:
* Put behind toggle.
* Create input field for port.
Misc cleanup:
* transcribe.py: Drop frames from audio capture thread instead of the
transcription thread. Doing it the other way would result in
occasional data loss.
Diffstat (limited to 'Scripts')
| -rw-r--r-- | Scripts/transcribe.py | 20 |
1 files changed, 14 insertions, 6 deletions
diff --git a/Scripts/transcribe.py b/Scripts/transcribe.py index 9711d15..51d7e0a 100644 --- a/Scripts/transcribe.py +++ b/Scripts/transcribe.py @@ -57,6 +57,7 @@ class AudioState: # Segment start time, end time, and text self.ranges_ls = [] self.frames = [] + self.drop_frames_till_i = -1 # Locks access to `text`. self.transcribe_lock = threading.Lock() @@ -131,6 +132,9 @@ def onAudioFramesAvailable( audio_state.CHUNK) if len(audio_state.frames) > max_frames: audio_state.frames = audio_state.frames[-1 * max_frames:] + if audio_state.drop_frames_till_i > 0: + audio_state.frames = audio_state.frames[audio_state.drop_frames_till_i:-1] + audio_state.drop_frames_till_i = -1 return (frames, pyaudio.paContinue) @@ -227,6 +231,7 @@ def transcribe(audio_state, model, frames, use_cpu: bool) -> typing.Tuple[str,st audio, beam_size = 5, language = audio_state.language, + temperature = [0.0], vad_filter = True, condition_on_previous_text = True, without_timestamps = False) @@ -244,17 +249,20 @@ def transcribe(audio_state, model, frames, use_cpu: bool) -> typing.Tuple[str,st for segment in ranges: first_segments.append(segment) break - if len(first_segments) >= 3: - c0 = first_segments[-3] + if len(first_segments) >= 5: + # Hack: require convergence across many frames to give the + # algorithm a longer buffer to work with. + c0 = first_segments[-1] c1 = first_segments[-2] - c2 = first_segments[-1] + c2 = first_segments[-3] + c3 = first_segments[-4] #print(f"c0: {c0}, c1: {c1}, c2: {c2}") - if c0 == c1 and c1 == c2: + if c0 == c1 and c1 == c2 and c2 == c3: # For simplicity, completely reset saved audio ranges. audio_state.ranges_ls = [] - committed_text = c2[2] + committed_text = c0[2] n_frames_to_drop = int(ceil(audio_state.RATE * c0[1])) - del audio_state.frames[0:n_frames_to_drop] + audio_state.drop_frames_till_i = n_frames_to_drop preview_text = "" for seg in ranges: |