app.py

import gradio as gr
import utils
import clean
import transcribe

css = """
.cursive-text {
    font-family: 'Brush Script MT', cursive;
}
"""

with gr.Blocks(theme="base", css=css) as demo:
    gr.Markdown("<center><h1> 🔊 Transcription <span class='cursive-text'>Delight</span> </h1></center>")
    gr.Markdown("### Step 1: Generate Raw Transcript")
    with gr.Row():
        with gr.Column():
            source = gr.Radio(label="Source type", choices=[("Audio", "audio"), ("Video", "video"), ("YouTube URL", "youtube")], value="audio")
            @gr.render(inputs=source)
            def show_source(s):
                if s == "audio":
                    source_component = gr.Audio(type="filepath")
                elif s == "video":
                    source_component = gr.Video()
                else:
                    source_component = gr.Textbox(placeholder="https://www.youtube.com/watch?v=44vi31hehw4")
                    preview = gr.HTML(label="Video preview")
                    source_component.change(utils.convert_to_embed_url, source_component, preview)

                    transcribe_btn.click(
                        utils.generate_audio,
                        [source, source_component],
                        [download_audio],
                        show_progress="minimal"
                    ).then(
                        transcribe.transcribe,
                        [download_audio],
                        [preliminary_transcript],
                    ).then(
                        lambda : [gr.Button(interactive=True), gr.CheckboxGroup(interactive=True)],
                        None,
                        [clean_btn, cleanup_options]
                    )

        with gr.Column():
            with gr.Row():
                transcribe_btn = gr.Button("Transcribe audio 📜", variant="primary")
                download_audio = gr.DownloadButton("Download .mp3 File 📥", interactive=False)
            preliminary_transcript = gr.Textbox(info="Raw transcript", lines=10, max_lines=10, show_copy_button=True, show_label=False, interactive=False)

    source.change(utils.transcribe_button, source, transcribe_btn)

    gr.Markdown("### Step 2: Clean with an LLM")
    with gr.Row():
        with gr.Column():
            cleanup_options = gr.CheckboxGroup(label="Cleanup Transcript with LLM", choices=["Remove typos", "Separate into paragraphs"])
            llm_prompt = gr.Textbox(label="LLM Prompt", visible=False, lines=3)
            cleanup_options.change(
                utils.generate_prompt,
                cleanup_options,
                llm_prompt
            )
            with gr.Row():
                clean_btn = gr.Button("Clean transcript ✨", variant="primary")            
                download_md = gr.DownloadButton("Download .md 📥", interactive=False)

        with gr.Column():
            final_transcript = gr.Markdown("*Final transcript will appear here*", height=400)

    clean_btn.click(
        clean.clean_transcript,
        [download_audio, cleanup_options, llm_prompt, preliminary_transcript],
        [final_transcript, download_md],
        show_progress="minimal"
    )
    
demo.launch()