Mirroristas/whisper
2
0
Fork 0
mirror of https://github.com/openai/whisper.git synced 2025-11-24 06:26:03 +00:00
Code Issues Packages Projects Releases Wiki Activity

Merge c1031a5787e7f21b789e9b84309d443d2fc7188a into c0d2f624c09dc18e709e37c2ad90c039a4eb72a2

Browse Source
This commit is contained in:
Hidetake Tanaka 2025-06-26 10:46:14 +05:30 committed by GitHub
commit 705b925bbf
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 7 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
whisper/transcribe.py
View File

@ -52,6 +52,7 @@ def transcribe(
append_punctuations: str = "\"'.。,，!！?？:：”)]}、", append_punctuations: str = "\"'.。,，!！?？:：”)]}、",
clip_timestamps: Union[str, List[float]] = "0", clip_timestamps: Union[str, List[float]] = "0",
hallucination_silence_threshold: Optional[float] = None, hallucination_silence_threshold: Optional[float] = None,
mel_spectrogram_device: Optional[Union[str, torch.device]] = None,
**decode_options, **decode_options,
): ):
""" """
@ -119,6 +120,9 @@ def transcribe(
When word_timestamps is True, skip silent periods longer than this threshold (in seconds) When word_timestamps is True, skip silent periods longer than this threshold (in seconds)
when a possible hallucination is detected when a possible hallucination is detected
mel_spectrogram_device: Optional[Union[str, torch.device]]
If given, the audio tensor is moved to this device before STFT
Returns Returns
------- -------
A dictionary containing the resulting text ("text") and segment-level details ("segments"), and A dictionary containing the resulting text ("text") and segment-level details ("segments"), and
@ -136,7 +140,9 @@ def transcribe(
decode_options["fp16"] = False decode_options["fp16"] = False
# Pad 30-seconds of silence to the input audio, for slicing # Pad 30-seconds of silence to the input audio, for slicing
mel = log_mel_spectrogram(audio, model.dims.n_mels, padding=N_SAMPLES) mel = log_mel_spectrogram(
audio, model.dims.n_mels, padding=N_SAMPLES, device=mel_spectrogram_device
)
content_frames = mel.shape[-1] - N_FRAMES content_frames = mel.shape[-1] - N_FRAMES
content_duration = float(content_frames * HOP_LENGTH / SAMPLE_RATE) content_duration = float(content_frames * HOP_LENGTH / SAMPLE_RATE)