Spaces:
Running
Running
| import base64 | |
| import os | |
| import time | |
| from dataclasses import dataclass | |
| from datetime import timedelta | |
| import banana_dev as banana | |
| import gradio as gr | |
| from loguru import logger | |
| from pydub import AudioSegment | |
| api_key = os.environ["BANANA_API_KEY"] | |
| model_key = os.environ["BANANA_MODEL_KEY"] | |
| password = os.environ["PASSWORD"] | |
| def transcribe(audio=None, url=None): | |
| if audio: | |
| audio_b64 = base64.b64encode(audio.export().read()).decode("ascii") | |
| payload = {"audio_b64": audio_b64} | |
| else: | |
| payload = {"url": url} | |
| response = banana.run(api_key, model_key, payload) | |
| print(response) | |
| if "error" in response: | |
| raise gr.Error(response["error"]) | |
| # TODO: not sure why response dict contains multiple model outputs | |
| return response["modelOutputs"][0] | |
| def run_demo(password, url, file_upload): | |
| if password not in [os.environ["PASSWORD"], os.environ["ROOT_PASSWORD"]]: | |
| raise gr.Error("Der Zugriffscode ist falsch.") | |
| if (url is not None) and (file_upload is not None): | |
| logger.warning( | |
| "Achtung: Sie haben sowohl eine URL angegeben als auch eine Datei hochgeladen." | |
| " Wir verwenden nur die Datei, die Sie hochgeladen haben." | |
| ) | |
| elif (url is None) and (file_upload is None): | |
| raise gr.Error( | |
| "Sie müssen entweder eine URL angeben oder eine Datei hochladen." | |
| ) | |
| start = time.time() | |
| if file_upload is not None: | |
| cutoff = None if password == os.environ["ROOT_PASSWORD"] else 60_000 | |
| audio = AudioSegment.from_file(file_upload, format="mp3")[:cutoff] | |
| transcription = transcribe(audio=audio, url=None) | |
| else: | |
| transcription = transcribe(audio=None, url=url) | |
| logger.info(f"transcription took {time.time()-start:.3f}s") | |
| return transcription["text"] | |
| demo = gr.Interface( | |
| fn=run_demo, | |
| inputs=[ | |
| # gr.Textbox(label="Email", type="email"), | |
| gr.Textbox(label="Zugriffscode (siehe oben)"), | |
| # gr.Audio(source="microphone", type="filepath", label="Aufnehmen"), | |
| gr.Textbox(label="URL (z.B. YouTube-Video, Dropbox-Datei, etc.)"), | |
| gr.Audio(source="upload", type="filepath", label="Datei hochladen"), | |
| ], | |
| outputs=gr.Textbox(label="Automatisches Transkript"), | |
| allow_flagging="never", | |
| css="footer {visibility: hidden} .meta-text {visibility: hidden}", | |
| ) | |
| demo.launch() | |