mirror of
https://github.com/chidiwilliams/buzz.git
synced 2026-03-14 14:45:46 +01:00
Will validate audio before transcribing (#1364)
This commit is contained in:
Raivis Dejus
GitHub
parent
0d446a9964
commit
a94d8fbd0d
2 changed files with 58 additions and 29 deletions
16
README.md
16
README.md
|
|
@ -13,7 +13,7 @@ OpenAI's [Whisper](https://github.com/openai/whisper).
|
|||
|
||||
[](https://GitHub.com/chidiwilliams/buzz/releases/)
|
||||
|
||||
|
||||
|
||||
|
||||
## Features
|
||||
- Transcribe audio and video files or Youtube links
|
||||
|
|
@ -91,12 +91,12 @@ For info on how to get latest development version with latest features and bug f
|
|||
### Screenshots
|
||||
|
||||
<div style="display: flex; flex-wrap: wrap;">
|
||||
<img alt="File import" src="share/screenshots/buzz-1-import.png" style="max-width: 18%; margin-right: 1%;" />
|
||||
<img alt="Main screen" src="share/screenshots/buzz-2-main_screen.png" style="max-width: 18%; margin-right: 1%; height:auto;" />
|
||||
<img alt="Preferences" src="share/screenshots/buzz-3-preferences.png" style="max-width: 18%; margin-right: 1%; height:auto;" />
|
||||
<img alt="Model preferences" src="share/screenshots/buzz-3.2-model-preferences.png" style="max-width: 18%; margin-right: 1%; height:auto;" />
|
||||
<img alt="Transcript" src="share/screenshots/buzz-4-transcript.png" style="max-width: 18%; margin-right: 1%; height:auto;" />
|
||||
<img alt="Live recording" src="share/screenshots/buzz-5-live_recording.png" style="max-width: 18%; margin-right: 1%; height:auto;" />
|
||||
<img alt="Resize" src="share/screenshots/buzz-6-resize.png" style="max-width: 18%;" />
|
||||
<img alt="File import" src="https://github.com/chidiwilliams/buzz/raw/main/share/screenshots/buzz-1-import.png" style="max-width: 18%; margin-right: 1%;" />
|
||||
<img alt="Main screen" src="https://github.com/chidiwilliams/buzz/raw/main/share/screenshots/buzz-2-main_screen.png" style="max-width: 18%; margin-right: 1%; height:auto;" />
|
||||
<img alt="Preferences" src="https://github.com/chidiwilliams/buzz/raw/main/share/screenshots/buzz-3-preferences.png" style="max-width: 18%; margin-right: 1%; height:auto;" />
|
||||
<img alt="Model preferences" src="https://github.com/chidiwilliams/buzz/raw/main/share/screenshots/buzz-3.2-model-preferences.png" style="max-width: 18%; margin-right: 1%; height:auto;" />
|
||||
<img alt="Transcript" src="https://github.com/chidiwilliams/buzz/raw/main/share/screenshots/buzz-4-transcript.png" style="max-width: 18%; margin-right: 1%; height:auto;" />
|
||||
<img alt="Live recording" src="https://github.com/chidiwilliams/buzz/raw/main/share/screenshots/buzz-5-live_recording.png" style="max-width: 18%; margin-right: 1%; height:auto;" />
|
||||
<img alt="Resize" src="https://github.com/chidiwilliams/buzz/raw/main/share/screenshots/buzz-6-resize.png" style="max-width: 18%;" />
|
||||
</div>
|
||||
|
||||
|
|
|
|||
|
|
@ -28,6 +28,7 @@ from buzz.transcriber.file_transcriber import FileTranscriber
|
|||
from buzz.transcriber.transcriber import FileTranscriptionTask, Segment, Task
|
||||
from buzz.transcriber.whisper_cpp import WhisperCpp
|
||||
|
||||
import av
|
||||
import faster_whisper
|
||||
import whisper
|
||||
import stable_whisper
|
||||
|
|
@ -36,6 +37,22 @@ from stable_whisper import WhisperResult
|
|||
PROGRESS_REGEX = re.compile(r"\d+(\.\d+)?%")
|
||||
|
||||
|
||||
def check_file_has_audio_stream(file_path: str) -> None:
|
||||
"""Check if a media file has at least one audio stream.
|
||||
|
||||
Raises:
|
||||
ValueError: If the file has no audio streams.
|
||||
"""
|
||||
try:
|
||||
with av.open(file_path) as container:
|
||||
if len(container.streams.audio) == 0:
|
||||
raise ValueError("No audio streams found")
|
||||
except av.error.InvalidDataError as e:
|
||||
raise ValueError(f"Invalid media file: {e}")
|
||||
except av.error.FileNotFoundError:
|
||||
raise ValueError("File not found")
|
||||
|
||||
|
||||
class WhisperFileTranscriber(FileTranscriber):
|
||||
"""WhisperFileTranscriber transcribes an audio file to text, writes the text to a file, and then opens the file
|
||||
using the default program for opening txt files."""
|
||||
|
|
@ -54,6 +71,7 @@ class WhisperFileTranscriber(FileTranscriber):
|
|||
self.stopped = False
|
||||
self.recv_pipe = None
|
||||
self.send_pipe = None
|
||||
self.error_message = None
|
||||
|
||||
def transcribe(self) -> List[Segment]:
|
||||
time_started = datetime.datetime.now()
|
||||
|
|
@ -119,7 +137,7 @@ class WhisperFileTranscriber(FileTranscriber):
|
|||
logging.debug("Whisper process was terminated (exit code: %s), treating as cancellation", self.current_process.exitcode)
|
||||
raise Exception("Transcription was canceled")
|
||||
else:
|
||||
raise Exception("Unknown error")
|
||||
raise Exception(self.error_message or "Unknown error")
|
||||
|
||||
return self.segments
|
||||
|
||||
|
|
@ -158,27 +176,36 @@ class WhisperFileTranscriber(FileTranscriber):
|
|||
subprocess.run = _patched_run
|
||||
subprocess.Popen = _PatchedPopen
|
||||
|
||||
with pipe_stderr(stderr_conn):
|
||||
if task.transcription_options.model.model_type == ModelType.WHISPER_CPP:
|
||||
segments = cls.transcribe_whisper_cpp(task)
|
||||
elif task.transcription_options.model.model_type == ModelType.HUGGING_FACE:
|
||||
sys.stderr.write("0%\n")
|
||||
segments = cls.transcribe_hugging_face(task)
|
||||
sys.stderr.write("100%\n")
|
||||
elif (
|
||||
task.transcription_options.model.model_type == ModelType.FASTER_WHISPER
|
||||
):
|
||||
segments = cls.transcribe_faster_whisper(task)
|
||||
elif task.transcription_options.model.model_type == ModelType.WHISPER:
|
||||
segments = cls.transcribe_openai_whisper(task)
|
||||
else:
|
||||
raise Exception(
|
||||
f"Invalid model type: {task.transcription_options.model.model_type}"
|
||||
)
|
||||
try:
|
||||
# Check if the file has audio streams before processing
|
||||
check_file_has_audio_stream(task.file_path)
|
||||
|
||||
segments_json = json.dumps(segments, ensure_ascii=True, default=vars)
|
||||
sys.stderr.write(f"segments = {segments_json}\n")
|
||||
sys.stderr.write(WhisperFileTranscriber.READ_LINE_THREAD_STOP_TOKEN + "\n")
|
||||
with pipe_stderr(stderr_conn):
|
||||
if task.transcription_options.model.model_type == ModelType.WHISPER_CPP:
|
||||
segments = cls.transcribe_whisper_cpp(task)
|
||||
elif task.transcription_options.model.model_type == ModelType.HUGGING_FACE:
|
||||
sys.stderr.write("0%\n")
|
||||
segments = cls.transcribe_hugging_face(task)
|
||||
sys.stderr.write("100%\n")
|
||||
elif (
|
||||
task.transcription_options.model.model_type == ModelType.FASTER_WHISPER
|
||||
):
|
||||
segments = cls.transcribe_faster_whisper(task)
|
||||
elif task.transcription_options.model.model_type == ModelType.WHISPER:
|
||||
segments = cls.transcribe_openai_whisper(task)
|
||||
else:
|
||||
raise Exception(
|
||||
f"Invalid model type: {task.transcription_options.model.model_type}"
|
||||
)
|
||||
|
||||
segments_json = json.dumps(segments, ensure_ascii=True, default=vars)
|
||||
sys.stderr.write(f"segments = {segments_json}\n")
|
||||
sys.stderr.write(WhisperFileTranscriber.READ_LINE_THREAD_STOP_TOKEN + "\n")
|
||||
except Exception as e:
|
||||
# Send error message back to the parent process
|
||||
stderr_conn.send(f"error = {str(e)}\n")
|
||||
stderr_conn.send(WhisperFileTranscriber.READ_LINE_THREAD_STOP_TOKEN + "\n")
|
||||
raise
|
||||
|
||||
@classmethod
|
||||
def transcribe_whisper_cpp(cls, task: FileTranscriptionTask) -> List[Segment]:
|
||||
|
|
@ -415,6 +442,8 @@ class WhisperFileTranscriber(FileTranscriber):
|
|||
for segment in segments_dict
|
||||
]
|
||||
self.segments = segments
|
||||
elif line.startswith("error = "):
|
||||
self.error_message = line[8:]
|
||||
else:
|
||||
try:
|
||||
match = PROGRESS_REGEX.search(line)
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue