Spaces:
Runtime error
Runtime error
| #uvicorn app789:app --host 0.0.0.0 --port 8000 --reload | |
| from fastapi import FastAPI, UploadFile, Form | |
| from fastapi.responses import HTMLResponse | |
| import librosa | |
| import io | |
| import json | |
| import requests | |
| import textwrap3 | |
| import whisper | |
| model = whisper.load_model("medium") | |
| app = FastAPI() | |
| # from faster_whisper import WhisperModel | |
| # model_size = "medium" | |
| # ts_model = WhisperModel(model_size, device="cpu", compute_type="int8") | |
| def read_root(): | |
| html_form = """ | |
| <html> | |
| <body> | |
| <h2>Audio Transcription</h2> | |
| <form action="/transcribe" method="post" enctype="multipart/form-data"> | |
| <label for="audio_file">Upload an audio file (MP3 or WAV):</label> | |
| <input type="file" id="audio_file" name="audio_file" accept=".mp3, .wav" required><br><br> | |
| <label for="language_select">Select Target Language:</label> | |
| <select id="language_select" name="tgt_lang"> | |
| <option value="fr_XX">French</option> | |
| <option value="es_XX">Spanish</option> | |
| <option value="de_DE">German</option> | |
| <option value="hi_IN">Hindi</option> | |
| <option value="en_XX">English</option> | |
| <option value="ja_XX">Japanese</option> | |
| <option value="ne_NP">Nepali</option> | |
| <option value="zh_CN">Chinese</option> | |
| <option value="pt_XX">Portuguese</option> | |
| <!-- Add more language options here --> | |
| </select><br><br> | |
| <input type="submit" value="Transcribe"> | |
| <input type="hidden" id="tgt_lang" name="tgt_lang" value="fr_XX"> | |
| </form> | |
| </body> | |
| </html> | |
| <script> | |
| document.getElementById("language_select").addEventListener("change", function () { | |
| var selectedLanguage = this.value; | |
| document.getElementById("tgt_lang").value = selectedLanguage; | |
| }); | |
| </script> | |
| """ | |
| return HTMLResponse(content=html_form, status_code=200) | |
| async def transcribe_audio(audio_file: UploadFile, tgt_lang: str = Form(...)): | |
| audio_data = await audio_file.read() | |
| audio_data, _ = librosa.load(io.BytesIO(audio_data), sr=16000) | |
| result = model.transcribe(audio_data, task = "translate") | |
| transcribed_text = result['text'] | |
| if tgt_lang == 'en_XX': | |
| return transcribed_text | |
| else: | |
| chunks = textwrap3.wrap(transcribed_text, 100) | |
| #segments, _ = ts_model.transcribe(audio_data, task="translate") | |
| # lst = [] | |
| # for segment in segments: | |
| # lst.append(segment.text) | |
| headers = {"Authorization": f"Bearer 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"} | |
| API_URL = "https://api-inference.huggingface.co/pipeline/translation/facebook/mbart-large-50-many-to-many-mmt" | |
| def query(payload): | |
| data = json.dumps(payload) | |
| response = requests.request("POST", API_URL, headers=headers, data=data) | |
| return json.loads(response.content.decode("utf-8")) | |
| translated_text = '' | |
| for i in chunks: | |
| result = query({"inputs": i, "parameters": {"src_lang": "en_XX", "tgt_lang": tgt_lang}}) | |
| translated_text = translated_text + result[0]['translation_text'] | |
| return translated_text | |