ExternalVendorCode/lollms
1
0
Fork 0
mirror of https://github.com/ParisNeo/lollms.git synced 2024-12-20 13:13:16 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

added whisper

Browse Source
This commit is contained in:
Saifeddine ALOUI 2024-02-01 23:31:14 +01:00
parent ec63c7be86
commit 86f8fab38f
2 changed files with 39 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

12
lollms/media.py
View File

@ -46,6 +46,12 @@ matplotlib.use('Agg')
if not PackageManager.check_package_installed("whisper"):
PackageManager.install_package("openai-whisper")
try:
import conda.cli
conda.cli.main("install", "conda-forge::ffmpeg", "-y")
except:
ASCIIColors.bright_red("Couldn't install ffmpeg. whisper won't work. Please install it manually")
import whisper
import socketio
@ -90,6 +96,7 @@ class AudioRecorder:
self.is_recording = False
self.start_time = time.time()
self.last_time = time.time()
self.whisper = whisper.load_model("base")
def audio_callback(self, indata, frames, time_, status):
volume_norm = np.linalg.norm(indata)*10
@ -120,6 +127,11 @@ class AudioRecorder:
write(self.filename, self.sample_rate, self.buffer)
self.lollmsCom.info(f"Saved to {self.filename}")
self.lollmsCom.info(f"Transcribing ... ")
result = self.whisper.transcribe(str(self.filename))
transcription_fn = str(self.filename)+".txt"
with open(transcription_fn, "w", encoding="utf-8") as f:
f.write(result["text"])
self.lollmsCom.info(f"File saved to {transcription_fn}")
def update_spectrogram(self):
f, t, Sxx = spectrogram(self.buffer[-30*self.sample_rate:], self.sample_rate)

28
lollms/personality.py
View File

@ -131,6 +131,9 @@ class AIPersonality:
self.installation_option = installation_option
# Whisper to transcribe audio
self.whisper = None
# First setup a default personality
# Version
self._version = pkg_resources.get_distribution('lollms').version
@ -762,7 +765,30 @@ Date: {{date}}
db_path = self.lollms_paths.personal_databases_path / "personalities" / self.name / "db.json"
db_path.parent.mkdir(parents=True, exist_ok=True)
path = Path(path)
if path.suffix in [".png",".jpg",".gif",".bmp",".webp"]:
if path.suffix in [".wav",".mp3"]:
if self.whisper is None:
if not PackageManager.check_package_installed("whisper"):
PackageManager.install_package("openai-whisper")
try:
import conda.cli
conda.cli.main("install", "conda-forge::ffmpeg", "-y")
except:
ASCIIColors.bright_red("Couldn't install ffmpeg. whisper won't work. Please install it manually")
import whisper
self.whisper = whisper.load_model("base")
self.info(f"Transcribing ... ")
self.step_start("Transcribing ... ")
result = self.whisper.transcribe(str(self.filename))
transcription_fn = str(path)+".txt"
with open(transcription_fn, "w", encoding="utf-8") as f:
f.write(result["text"])
self.info(f"File saved to {transcription_fn}")
self.full(result["text"])
self.step_end("Transcribing ... ")
elif path.suffix in [".png",".jpg",".gif",".bmp",".webp"]:
if self.callback:
try:
pth = str(path).replace("\\","/").split('/')