Spaces:
Sleeping
Sleeping
File size: 1,709 Bytes
93d50a2 a5434d9 93d50a2 a5434d9 93d50a2 a5434d9 60c7aa2 a5434d9 60c7aa2 a5434d9 60c7aa2 a5434d9 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 |
---
title: Talklas 2
emoji: 🚀
colorFrom: blue
colorTo: green
sdk: docker
app_port: 8000
health_check_path: /health
---
# Talklas API
This FastAPI app is deployed on Hugging Face Spaces for audio transcription, translation, and text-to-speech (TTS). The application loads all necessary models at startup and updates each model on demand, including a feature for detecting inappropriate language. It includes the following endpoints:
- `/`: Returns a simple health check response.
- `/health`: Health check endpoint for Hugging Face Spaces.
- `/update-languages`: Updates the source and target languages for STT and TTS models.
- `/translate-text`: Translates text and converts it to speech.
- `/translate-audio`: Transcribes audio, translates the text, and converts the translated text to speech. Includes speech detection to handle silent audio gracefully.
- `/text-to-speech`: This will generate a speech from the text and target language given to the server
## Features
- **Speech Detection**: The `/translate-audio` endpoint detects if the audio is silent (no speech) and returns a user-friendly response.
- **Transcription (STT)**: Uses Whisper for English and Tagalog source language and MMS for other Philippine languages to transcribe audio.
- **Translation (MT)**: Uses the NLLB-200 model to translate text between supported languages.
- **Text-to-Speech (TTS)**: Uses MMS-TTS models to convert translated text to speech.
## Supported Languages
- English
- Tagalog
- Cebuano
- Ilocano
- Waray
- Pangasinan
## Deployment
This app uses a `Dockerfile` to deploy a FastAPI app with Uvicorn. The health check path is set to `/health` to ensure Hugging Face Spaces can verify the app is running. |