Update app.py

app.py

@@ -1,47 +1,104 @@
-import gradio as gr
+import torch
 from transformers import pipeline
+from transformers.pipelines.audio_utils import ffmpeg_read
+import gradio as gr
 
-# Load the ASR model using the Hugging Face pipeline
-model_id = "riteshkr/whisper-large-v3-quantized"
-pipe = pipeline("automatic-speech-recognition", model=model_id)
-
-# Define the transcription function
-def transcribe_speech(filepath):
-    output = pipe(
-        filepath,
-        max_new_tokens=256,
-        generate_kwargs={
-            "task": "transcribe",
-            "language": "english",
-        },
-        chunk_length_s=30,
-        batch_size=8,
-    )
-    return output["text"]
+MODEL_NAME = "riteshkr/whisper-large-v3-quantized"
+BATCH_SIZE = 8
+
+device = 0 if torch.cuda.is_available() else "cpu"
+
+pipe = pipeline(
+    task="automatic-speech-recognition",
+    model=MODEL_NAME,
+    chunk_length_s=30,
+    device=device,
+)
+
+
+# Copied from https://github.com/openai/whisper/blob/c09a7ae299c4c34c5839a76380ae407e7d785914/whisper/utils.py#L50
+def format_timestamp(seconds: float, always_include_hours: bool = False, decimal_marker: str = "."):
+    if seconds is not None:
+        milliseconds = round(seconds * 1000.0)
+
+        hours = milliseconds // 3_600_000
+        milliseconds -= hours * 3_600_000
+
+        minutes = milliseconds // 60_000
+        milliseconds -= minutes * 60_000
+
+        seconds = milliseconds // 1_000
+        milliseconds -= seconds * 1_000
+
+        hours_marker = f"{hours:02d}:" if always_include_hours or hours > 0 else ""
+        return f"{hours_marker}{minutes:02d}:{seconds:02d}{decimal_marker}{milliseconds:03d}"
+    else:
+        # we have a malformed timestamp so just return it as is
+        return seconds
+
+
+def transcribe(file, task, return_timestamps):
+    outputs = pipe(file, batch_size=BATCH_SIZE, generate_kwargs={"task": task}, return_timestamps=return_timestamps)
+    text = outputs["text"]
+    if return_timestamps:
+        timestamps = outputs["chunks"]
+        timestamps = [
+            f"[{format_timestamp(chunk['timestamp'][0])} -> {format_timestamp(chunk['timestamp'][1])}] {chunk['text']}"
+            for chunk in timestamps
+        ]
+        text = "\n".join(str(feature) for feature in timestamps)
+    return text
+
+
+demo = gr.Blocks()
 
-# Define the Gradio interface for microphone input
 mic_transcribe = gr.Interface(
-    fn=transcribe_speech,
-    inputs=gr.Audio(sources="microphone", type="filepath"),
-    outputs=gr.Textbox(),
+    fn=transcribe,
+    inputs=[
+        gr.Audio(sources="microphone", type="filepath"),
+        gr.Radio(["transcribe", "translate"], label="Task"),
+        gr.Checkbox(label="Return timestamps"),
+    ],
+    outputs="text",
+    layout="horizontal",
+    theme="huggingface",
+    title="Transcribe Audio",
+    description=(
+        "Transcribe long-form microphone or audio inputs with the click of a button! Demo uses the"
+        f" checkpoint [{MODEL_NAME}](https://huggingface.co/{MODEL_NAME}) and 🤗 Transformers to transcribe audio files"
+        " of arbitrary length."
+    ),
+    allow_flagging="never",
 )
 
-# Define the Gradio interface for file upload input
 file_transcribe = gr.Interface(
-    fn=transcribe_speech,
-    inputs=gr.Audio(sources="upload", type="filepath"),
-    outputs=gr.Textbox(),
+    fn=transcribe,
+    inputs=[
+        gr.Audio(sources="upload", label="Audio file", type="filepath"),
+        gr.Radio(["transcribe", "translate"], label="Task"),
+        gr.Checkbox(label="Return timestamps"),
+    ],
+    outputs="text",
+    layout="horizontal",
+    theme="huggingface",
+    title="Transcribe Audio",
+    description=(
+        "Transcribe long-form microphone or audio inputs with the click of a button! Demo uses the"
+        f" checkpoint [{MODEL_NAME}](https://huggingface.co/{MODEL_NAME}) and 🤗 Transformers to transcribe audio files"
+        " of arbitrary length."
+    ),
+    examples=[
+        ["./example.flac", "transcribe", False],
+        ["./example.flac", "transcribe", True],
+    ],
+    cache_examples=True,
+    allow_flagging="never",
 )
 
-# Creating the tabbed layout using Blocks
-demo = gr.Blocks()
-
 with demo:
     gr.TabbedInterface(
-        [mic_transcribe, file_transcribe],
-        ["Transcribe Microphone", "Transcribe Audio File"],
+        [mic_transcribe, file_transcribe], 
+        ["Transcribe Microphone", "Transcribe Audio File"]
     )
 
-# Launch the app with debugging enabled
-if __name__ == "__main__":
-    demo.launch(debug=True, share=True)
+demo.launch(enable_queue=True)