app.py

import nemo.collections.asr as nemo_asr
import gradio as gr




asr_model = nemo_asr.models.EncDecHybridRNNTCTCBPEModel.from_pretrained("theodotus/stt_ua_fastconformer_hybrid_large_pc", map_location="cpu")




def process_file(in_filename: str,):
    if in_filename is None or in_filename == "":
        return "Error: No file"

    transcript = asr_model.transcribe(paths2audio_files = [in_filename])[0][0]


    return transcript




demo = gr.Blocks()

with demo:
    with gr.Tabs():
        with gr.TabItem("Upload from disk"):
            uploaded_file = gr.Audio(
                source="upload",  # Choose between "microphone", "upload"
                type="filepath",
                optional=False,
                label="Upload from disk",
            )
            upload_button = gr.Button("Submit for recognition")
            uploaded_output = gr.Textbox(label="Recognized speech from uploaded file")

        with gr.TabItem("Record from microphone"):
            microphone = gr.Audio(
                source="microphone",  # Choose between "microphone", "upload"
                type="filepath",
                optional=False,
                label="Record from microphone",
            )

            record_button = gr.Button("Submit for recognition")
            recorded_output = gr.Textbox(label="Recognized speech from recordings")

        upload_button.click(
            process_file,
            inputs=[
                uploaded_file,
            ],
            outputs=[uploaded_output],
        )

        record_button.click(
            process_file,
            inputs=[
                microphone,
            ],
            outputs=[recorded_output],
        )


if __name__ == "__main__":
    demo.launch()