Create app.py

app.py

@@ -0,0 +1,57 @@
+import gradio as gr
+import torch
+from transformers import AutoModelForSpeechSeq2Seq, AutoProcessor, pipeline
+from datasets import load_dataset
+import soundfile as sf
+
+# Check if CUDA is available and set the device
+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 processor
+model_id = "openai/whisper-large-v3"
+model = AutoModelForSpeechSeq2Seq.from_pretrained(
+    model_id, torch_dtype=torch_dtype, low_cpu_mem_usage=True, use_safetensors=True
+).to(device)
+processor = AutoProcessor.from_pretrained(model_id)
+
+# Define the ASR pipeline
+asr_pipeline = 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,
+)
+
+# Function to process audio in chunks and return the combined text
+def process_audio(file_info):
+    path = file_info["path"]
+    audio_stream = sf.SoundFile(path, 'r')
+    results = []
+    while True:
+        data = audio_stream.read(dtype='float32')
+        if len(data) == 0:
+            break
+        result = asr_pipeline(data)
+        results.append(result)
+    audio_stream.close()
+    combined_text = " ".join([r["text"] for r in results])
+    return combined_text
+
+# Create the Gradio interface
+iface = gr.Interface(
+    fn=process_audio,
+    inputs=gr.inputs.Audio(source="upload", type="file", label="Upload your audio file"),
+    outputs="text",
+    title="👋🏻Welcome To 🙋🏻‍♂️Patrick's Whisper🌬️",
+    description="Upload a large audio file to transcribe it into text using [Whisper3Large](https://huggingface.co/openai/whisper-large-v3) !",
+)
+
+# Launch the application
+iface.launch()