BeastGokul commited on
Commit
b5712d1
·
verified ·
1 Parent(s): f6b0982

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +3 -8
app.py CHANGED
@@ -39,12 +39,6 @@ model_name = "BeastGokul/Nika-1.5B"
39
  llm_tokenizer = AutoTokenizer.from_pretrained(model_name)
40
  llm_model = AutoModelForCausalLM.from_pretrained(model_name)
41
 
42
- # Option 2: OpenAI Whisper for speech recognition
43
- whisper_processor = AutoProcessor.from_pretrained("openai/whisper-large-v3")
44
- whisper_model = AutoModelForSpeechSeq2Seq.from_pretrained("openai/whisper-large-v3")
45
-
46
- # Option 3: Wav2Vec2 for phoneme-level analysis
47
-
48
  # Automatically use GPU if available
49
  device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
50
 
@@ -120,8 +114,9 @@ def save_audio(audio, user_id="default"):
120
 
121
  # Audio processing and phonetic analysis
122
  def transcribe_with_whisper(audio_path):
123
- """Transcribe audio using OpenAI's Whisper model"""
124
- result = whisper_model.transcribe(audio_path)
 
125
  return result["text"]
126
 
127
  def extract_phonemes(text):
 
39
  llm_tokenizer = AutoTokenizer.from_pretrained(model_name)
40
  llm_model = AutoModelForCausalLM.from_pretrained(model_name)
41
 
 
 
 
 
 
 
42
  # Automatically use GPU if available
43
  device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
44
 
 
114
 
115
  # Audio processing and phonetic analysis
116
  def transcribe_with_whisper(audio_path):
117
+ """Transcribe audio using Hugging Face's pipeline for Whisper"""
118
+ asr_pipeline = pipeline("automatic-speech-recognition", model="openai/whisper-large-v3")
119
+ result = asr_pipeline(audio_path)
120
  return result["text"]
121
 
122
  def extract_phonemes(text):