AI test version super initial commit

Dockerfile

@@ -0,0 +1,14 @@
+FROM python:3.10-slim
+
+WORKDIR /code
+
+COPY requirements.txt .
+RUN pip install --no-cache-dir -r requirements.txt
+
+COPY app/ ./app
+
+WORKDIR /code/app
+
+EXPOSE 7860
+
+CMD ["uvicorn", "main:app", "--host", "0.0.0.0", "--port", "7860"]

app/config.py

@@ -0,0 +1,36 @@
+import os
+from dataclasses import dataclass, field
+from typing import Dict
+
+
+
+
+ELEVENLABS_API_KEY = "ELEVENLABS_API_KEY"
+
+ELEVENLABS_BASE_URL = "https://api.elevenlabs.io/v1"
+
+
+MODEL_ID = "eleven_v3"
+
+
+VOICE_IDS = {
+    "ig": "QLniWkGYsJa91mXrxl3c",   # Igbo
+    "yo": "x86DtpnPPuq2BpEiKPRy",   # Yoruba
+    "ha": "65blg8G9v7ZRWectRxK8",   # Hausa
+    "en": "65blg8G9v7ZRWectRxK8",   # English
+}
+
+DEFAULT_LANGUAGE = "en"
+
+
+CONFIDENCE_THRESHOLD = 0.5
+
+FASTTEXT_REPO_ID = "facebook/fasttext-language-identification"
+FASTTEXT_FILENAME = "model.bin"
+
+LABEL_TO_VOICE = {
+    "ibo_Latn": "ig",
+    "yor_Latn": "yo",
+    "hau_Latn": "ha",
+    "eng_Latn": "en",
+}

app/main.py

@@ -0,0 +1,75 @@
+import os
+import requests
+from fastapi import FastAPI, HTTPException
+from fastapi.responses import Response
+from pydantic import BaseModel
+
+from app.config import (
+    ELEVENLABS_API_KEY,
+    ELEVENLABS_BASE_URL,
+    MODEL_ID,
+    VOICE_IDS,
+    DEFAULT_LANGUAGE,
+)
+   
+from app.voice_languag_detector import detect_language
+
+app = FastAPI(title="Africa TTS - Auto Language Voice Cloning")
+
+
+class TTSRequest(BaseModel):
+    text: str
+    stability: float = 0.5
+    similarity_boost: float = 0.75
+
+
+@app.get("/")
+def root():
+    return {"status": "ok", "message": "Africa TTS API is running"}
+
+
+@app.post("/tts")
+def text_to_speech(request: TTSRequest):
+    if not request.text or not request.text.strip():
+        raise HTTPException(status_code=400, detail="Text cannot be empty")
+
+    if not ELEVENLABS_API_KEY:
+        raise HTTPException(status_code=500, detail="ELEVENLABS_API_KEY not set")
+
+    lang_code, confidence = detect_language(request.text)
+    voice_id = VOICE_IDS.get(lang_code, VOICE_IDS[DEFAULT_LANGUAGE])
+
+    url = f"{ELEVENLABS_BASE_URL}/text-to-speech/{voice_id}"
+
+    headers = {
+        "xi-api-key": ELEVENLABS_API_KEY,
+        "Content-Type": "application/json",
+        "Accept": "audio/mpeg",
+    }
+
+    payload = {
+        "text": request.text,
+        "model_id": MODEL_ID,
+        "voice_settings": {
+            "stability": request.stability,
+            "similarity_boost": request.similarity_boost,
+        },
+    }
+
+    response = requests.post(url, json=payload, headers=headers)
+
+    if response.status_code != 200:
+        raise HTTPException(
+            status_code=response.status_code,
+            detail=f"ElevenLabs API error: {response.text}",
+        )
+
+    return Response(
+      content = response.content, 
+      media_type = "audio/mpeg",
+      headers = {
+            "X-Detected-Language": lang_code,
+            "X-Detection-confidence": str(round(confidence,4)),
+            "X-Voice-Id-Used": voice_id,
+      }
+    )

app/voice_languag_detector.py

@@ -0,0 +1,45 @@
+import fasttext
+from huggingface_hub import hf_hub_download
+
+from app.config import (
+    FASTTEXT_REPO_ID,
+    FASTTEXT_FILENAME,
+    CONFIDENCE_THRESHOLD,
+    DEFAULT_LANGUAGE,
+    LABEL_TO_VOICE,
+)
+
+fasttext.FastText.eprint = lambda x: None  # silence warnings
+
+
+class LanguageDetector:
+    def __init__(self):
+        model_path = hf_hub_download(
+            repo_id=FASTTEXT_REPO_ID,
+            filename=FASTTEXT_FILENAME,
+        )
+        self.model = fasttext.load_model(model_path)
+
+    def detect(self, text: str):
+        clean_text = text.replace("\n", " ").strip()
+        if not clean_text:
+            return DEFAULT_LANGUAGE, 0.0
+
+        labels, confidences = self.model.predict(clean_text, k=1)
+        raw_label = labels[0].replace("__label__", "")
+        confidence = float(confidences[0])
+
+        lang_code = LABEL_TO_VOICE.get(raw_label)
+
+        if lang_code is not None and confidence >= CONFIDENCE_THRESHOLD:
+            return lang_code, confidence
+
+        return DEFAULT_LANGUAGE, confidence
+
+
+detector = LanguageDetector()
+
+
+def detect_language(text: str):
+    """Returns (lang_code, confidence)"""
+    return detector.detect(text)

classifier_model/voice_language_detector.py

@@ -0,0 +1,41 @@
+import os
+import urllib.request
+import fasttext
+
+from app.config import config
+    
+
+
+fasttext.FastText.eprint = lambda x: None
+
+
+class LanguageDetector:
+    def __init__(self):
+        if not os.path.exists(FASTTEXT_MODEL_PATH):
+            print("Downloading fastText language ID model")
+            urllib.request.urlretrieve(FASTTEXT_MODEL_URL, FASTTEXT_MODEL_PATH)
+        self.model = fasttext.load_model(FASTTEXT_MODEL_PATH)
+
+    def detect(self, text: str):
+        clean_text = text.replace("\n", " ").strip()
+        if not clean_text:
+            return DEFAULT_LANGUAGE, 0.0
+
+        labels, confidences = self.model.predict(clean_text, k=1)
+        lang_code = labels[0].replace("__label__", "")
+        confidence = float(confidences[0])
+
+        
+        if lang_code in VOICE_IDS and confidence >= CONFIDENCE_THRESHOLD:
+            return lang_code, confidence
+
+        return DEFAULT_LANGUAGE, confidence
+
+
+
+detector = LanguageDetector()
+
+
+def detect_language(text: str):
+    """Returns (lang_code, confidence)"""
+    return detector.detect(text)

lid.176.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7e69ec5451bc261cc7844e49e4792a85d7f09c06789ec800fc4a44aec362764e
+size 131266198

requirements.txt

@@ -0,0 +1,6 @@
+fastapi
+uvicorn[standard]
+requests
+fasttext-wheel
+pydantic 
+numpy<2