app.py

import os
import gradio as gr
import json
import asyncio
import aiohttp
import requests
from tqdm.asyncio import tqdm

API_BASE_URL = os.environ.get("DUBPRO_API_BASE_URL")

def fetch_data_from_db(_id, details="GET_ADVANCE_DETAILS"):
    API_URL = f"{API_BASE_URL}/video-service/videos/{_id}/get_details"
    headers = {
        "Accept": "*/*",
        "Authorization": os.environ.get("DUBPRO_API_HEADER_KEY")
    }
    params = {"type": details}
    response = requests.get(API_URL, headers=headers, params=params)
    if response.status_code == 200:
        data = response.json()
        return data["data"]
    else:
        raise Exception(response.text)


async def generate_audio(session, video_id, segment_id, text, progress_bar):
    payload = {"type": "EDIT_TEXT", "updatedText": text}
    headers = {"Authorization": os.environ.get("DUBPRO_API_HEADER_KEY")}
    API_URL = f"{API_BASE_URL}/video-service/videos/{video_id}/segment/{segment_id}"

    async with session.put(API_URL, headers=headers, json=payload) as response:
        if response.status != 200:
            print(segment_id, text, await response.text())
    progress_bar.update(1)


async def generate_audios(video_id, segments=[]):
    async with aiohttp.ClientSession() as session:
        progress_bar = tqdm(total=len(segments), desc="Generating Audios", unit="segment")
        tasks = [
            generate_audio(session, video_id, segment["id"], segment["updatedText"], progress_bar) 
            for segment in segments
        ]
        await asyncio.gather(*tasks)
        progress_bar.close()


def perform_generation(video_id, gen_choice, progress_bar=gr.Progress(track_tqdm=True)):
    data = fetch_data_from_db(video_id)
    segments = data["targetTranscriptData"]
    if gen_choice == "Generate only non-generated segments":
        segments = [s for s in segments if s["audioUrl"] is None or "updated" not in s["audioUrl"]]
    elif gen_choice == "Re-generate generated only segments":
        segments = [s for s in segments if s["audioUrl"] is not None and "updated" in s["audioUrl"]]
    asyncio.run(generate_audios(video_id, segments=segments))
    return "Done"


async def update_tempo(session, video_id, segment_id, tempo):
    payload = {"type": "EDIT_SPEED", "tempo": tempo}
    headers = {"Authorization": os.environ.get("DUBPRO_API_HEADER_KEY")}
    API_URL = f"{API_BASE_URL}/video-service/videos/{video_id}/segment/{segment_id}"

    async with session.put(API_URL, headers=headers, json=payload) as response:
        if response.status != 200:
            print(await response.text())


async def process_segment(session, video_id, segment, progress_bar):
    WARNING_BOUNDARY = 0.1
    duration = round(segment["endTimestamp"] - segment["startTimestamp"], 3)
    if WARNING_BOUNDARY <= abs(duration - segment["audioDuration"]) <= (WARNING_BOUNDARY + 2.5):
        if duration - segment["audioDuration"] > 0:
            # Gap case
            tempo = round(segment["audioDuration"] / duration, 3)
        else:
            tempo = round(segment["audioDuration"] / (duration + 0.1), 3)
        await update_tempo(session, video_id, segment["id"], tempo)
    progress_bar.update(1)


async def update_tempos(video_id, segments=[]):
    async with aiohttp.ClientSession() as session:
        progress_bar = tqdm(total=len(segments), desc="Processing Segments", unit="segment")

        tasks = [
            process_segment(session, video_id, segment, progress_bar) 
            for segment in segments
        ]
        await asyncio.gather(*tasks)
        progress_bar.close()


def perform_tempo_adjustment(video_id, gen_choice, progress_bar=gr.Progress(track_tqdm=True)):
    data = fetch_data_from_db(video_id)
    segments = data["targetTranscriptData"]
    if gen_choice == "Generate only non-generated segments":
        segments = [s for s in segments if s["audioUrl"] is None or "updated" not in s["audioUrl"]]
    elif gen_choice == "Re-generate generated only segments":
        segments = [s for s in segments if s["audioUrl"] is not None and "updated" in s["audioUrl"]]
    asyncio.run(update_tempos(video_id, segments=segments))
    return "Done"


def check_details(video_id):
    data = fetch_data_from_db(video_id)
    segments = data["targetTranscriptData"]
    results = json.dumps(segments[0]) + "\n\n" + json.dumps(segments[-1])
    return results, gr.update(visible=True)


with gr.Blocks() as demo:
    gr.Markdown("# Generate Audio")
    video_id_txt = gr.Text(label="Video ID")
    gen_choice_radio = gr.Radio(["Generate only non-generated segments", "Re-generate generated only segments", "Generate all the segments"], label="Audio generation style", value="Generate only non-generated segments")
    check_settings_btn = gr.Button("Check settings")
    logs_txt = gr.Textbox(label="Logs")
    generate_btn = gr.Button("Generate audio", visible=False)

    check_settings_btn.click(check_details, video_id_txt, [logs_txt, generate_btn])
    generate_btn.click(perform_generation, [video_id_txt, gen_choice_radio], logs_txt)

    gr.Markdown("## Update Tempo [SAH]")
    update_tempo_btn = gr.Button("Adjust tempo")
    update_tempo_btn.click(perform_tempo_adjustment, [video_id_txt, gen_choice_radio], logs_txt)
    

if __name__=="__main__":
    demo.queue().launch(auth=(os.environ.get("GRADIO_USERNAME"), os.environ.get("GRADIO_PASSWORD")))