app.py

import gradio as gr
from tone import StreamingCTCPipeline, read_audio

# Load model once at startup
pipe = StreamingCTCPipeline.from_hugging_face()

def transcribe(audio_path):
    if audio_path is None:
        return "Please upload an audio file."
    try:
        audio = read_audio(audio_path)
        result = pipe.forward_offline(audio)
        # result: list of TextPhrase with text, start_time, end_time
        return "\n".join([x.text for x in result])
    except Exception as e:
        return f"Error: {str(e)}"

with gr.Blocks(title="T-one ASR Demo") as demo:
    gr.Markdown("# T-one Automatic Speech Recognition Demo")
    gr.Markdown("Upload an audio file to get real-time transcription using the t-tech/T-one model.")
    audio_input = gr.Audio(label="Upload Audio File", type="filepath")
    text_output = gr.Textbox(label="Transcription", placeholder="Transcribed text will appear here...", lines=5)
    audio_input.change(transcribe, inputs=audio_input, outputs=text_output)

demo.launch()