| import gradio as gr |
| import spaces |
| from faster_whisper import WhisperModel |
|
|
| |
| MODEL_ID = "collabora/faster-whisper-large-v2-hindi" |
|
|
| @spaces.GPU(duration=120) |
| def transcribe_hindi(audio): |
| if audio is None: |
| return "कृपया ऑडियो प्रदान करें।", None |
| |
| |
| model = WhisperModel(MODEL_ID, device="cuda", compute_type="float16") |
| |
| |
| segments, info = model.transcribe(audio, language="hi", beam_size=5) |
| full_text = "".join([segment.text for segment in segments]) |
| |
| |
| file_path = "transcription.txt" |
| with open(file_path, "w", encoding="utf-8") as f: |
| f.write(full_text.strip()) |
| |
| return full_text.strip(), file_path |
|
|
| custom_css = """ |
| footer {visibility: hidden} |
| .gradio-container {background-color: #fcfcfc} |
| #header {text-align: center; margin-bottom: 20px} |
| """ |
|
|
| with gr.Blocks(title="IndicWhisper Collabora GPU") as demo: |
| gr.HTML("<div id='header'><h1>🎙️ Hindi Whisper (Collabora - faster_whisper)</h1></div>") |
| |
| with gr.Row(): |
| with gr.Column(): |
| audio_input = gr.Audio( |
| sources=["microphone", "upload"], |
| type="filepath", |
| label="ऑडियो रिकॉर्ड करें या अपलोड करें" |
| ) |
| submit_btn = gr.Button("अनुवाद करें (Transcribe)", variant="primary") |
| |
| with gr.Column(): |
| output_text = gr.Textbox( |
| label="Transcription Output", |
| lines=10, |
| placeholder="आपका टेक्स्ट यहाँ दिखाई देगा..." |
| ) |
| download_file = gr.File( |
| label="टैक्स्ट फ़ाइल डाउनलोड करें", |
| visible=True |
| ) |
| |
| gr.Markdown(""" |
| --- |
| **सुझाव:** यहाँ से प्राप्त आउटपुट को कॉपी करें और अपने **Gemini Gem** में पेस्ट करें ताकि **पञ्चमाक्षर नियमों** (ङ्, ञ्, ण्, न्, म्) के अनुसार शुद्धिकरण किया जा सके। |
| """) |
|
|
| submit_btn.click( |
| fn=transcribe_hindi, |
| inputs=audio_input, |
| outputs=[output_text, download_file] |
| ) |
|
|
| if __name__ == "__main__": |
| demo.launch(css=custom_css) |
