Spaces:
Sleeping
Sleeping
File size: 1,744 Bytes
c8d4343 77e5ef6 c8d4343 77e5ef6 36cffb4 77e5ef6 c8d4343 b0b7514 1651123 c8d4343 1651123 c8d4343 77e5ef6 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 |
import gradio as gr
import torch
from transformers import AutoModelForSpeechSeq2Seq, AutoProcessor, pipeline
# Check if CUDA is available, and choose device accordingly
device = "cuda:0" if torch.cuda.is_available() else "cpu"
torch_dtype = torch.float16 if torch.cuda.is_available() else torch.float32
# Load the model and tokenizer
model_id = "openai/whisper-large-v3"
model = AutoModelForSpeechSeq2Seq.from_pretrained(
model_id, torch_dtype=torch_dtype, use_safetensors=True
)
model.to(device)
processor = AutoProcessor.from_pretrained(model_id)
# Define a function to transcribe audio
pipe = pipeline(
"automatic-speech-recognition",
model=model,
tokenizer=processor.tokenizer,
feature_extractor=processor.feature_extractor,
max_new_tokens=128,
chunk_length_s=30,
batch_size=16,
return_timestamps=True,
torch_dtype=torch_dtype,
device=device,
)
def transcribe_audio(audio_file):
# Check if audio file is None
#if audio_file is None:
# raise ValueError("Input audio file is None.")
# Use the pipeline to transcribe audio
result = pipe(audio_file, generate_kwargs={"language": "english"})
transcribed_text = result["text"]
return transcribed_text
# Create a Gradio interface
audio_input = gr.Audio(label="Upload Audio", type="filepath")
output_text = gr.Textbox(label="Transcribed Text")
# Instantiate the Gradio interface
app = gr.Interface(
fn=transcribe_audio,
inputs=audio_input,
outputs=output_text,
title="Audio Transcription with Whisper Model",
description="Upload an audio file to transcribe it into text using the Whisper model.",
theme="compact"
)
# Launch the Gradio interface
app.launch(debug=True, inline=False)
|