Delete app.py

from transformers import pipeline
import gradio as gr
import librosa
import numpy as np

from transformers import WhisperProcessor, WhisperForConditionalGeneration, pipeline

processor = WhisperProcessor.from_pretrained("https://huggingface.co/spaces/akadriu/shqip_whisper")
model = WhisperForConditionalGeneration.from_pretrained("https://huggingface.co/spaces/akadriu/shqip_whisper")

def transcribe(audio):
audio_input, _ = librosa.load(audio, sr=16000)
input_features = processor(audio_input, sampling_rate=16000, return_tensors="pt").input_features
predicted_ids = model.generate(input_features)
transcription = processor.batch_decode(predicted_ids, skip_special_tokens=True)
text = transcription
return text

iface = gr.Interface(
fn=transcribe,
inputs=gr.Audio(source="microphone", type="filepath"),
outputs="text",
title="Whisper Medium Shqip",
description="Realtime demo for Sq speech recognition using a fine-tuned Whisper medium model.",
)

iface.launch()

app.py

@@ -1,45 +0,0 @@
-from transformers import pipeline
-import gradio as gr
-import librosa
-import numpy as np
-import os
-
-from transformers import WhisperProcessor, WhisperForConditionalGeneration
-
-hf_token = os.getenv("HUGGINGFACE_HUB_TOKEN")
-processor = WhisperProcessor.from_pretrained("akadriu/whisper-medium-sq", token=hf_token)
-model = WhisperForConditionalGeneration.from_pretrained("akadriu/whisper-medium-sq", token=hf_token)
-
-def transcribe(audio):
-    # If audio is a tuple, extract the NumPy array and the sampling rate
-    if isinstance(audio, tuple):
-        audio_input, sr = audio
-        # Save the audio back to a file for playback
-        audio_file = "temp.wav"
-        librosa.output.write_wav(audio_file, audio_input, sr)
-    else:
-        # Otherwise, load the file from the path
-        audio_file = audio
-        audio_input, sr = librosa.load(audio_file, sr=16000)
-    
-    # Ensure the sample rate is what the processor expects
-    if sr != 16000:
-        audio_input = librosa.resample(audio_input, orig_sr=sr, target_sr=16000)
-    
-    # Process the audio and generate the transcription
-    input_features = processor(audio_input, sampling_rate=16000, return_tensors="pt").input_features
-    predicted_ids = model.generate(input_features)
-    transcription = processor.batch_decode(predicted_ids, skip_special_tokens=True)
-    text = transcription[0]  # Decode returns a list
-    
-    return audio_file, text
-
-iface = gr.Interface(
-    fn=transcribe,
-    inputs=gr.Audio(),  # Audio input
-    outputs=[gr.Audio(), "text"],  # Return the audio file and transcription text
-    title="Whisper Medium Shqip",
-    description="Realtime demo for Sq speech recognition using a fine-tuned Whisper medium model.",
-)
-
-iface.launch(share=True)