Update app.py

app.py

@@ -1,30 +1,47 @@
+from fastapi import FastAPI, Request
+from fastapi.responses import JSONResponse
 from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
 import torch
 import librosa
-import gradio as gr
+import tempfile
+import base64
+import requests
 
-# Cargamos el modelo de guaraní
+app = FastAPI()
+
+# Cargar modelo Guaraní
 model_name = "ivangtorre/wav2vec2-xlsr-300m-guarani"
 processor = Wav2Vec2Processor.from_pretrained(model_name)
 model = Wav2Vec2ForCTC.from_pretrained(model_name)
 
-# Transcripción
-def transcribir(audio):
-    audio_data, _ = librosa.load(audio, sr=16000)
-    inputs = processor(audio_data, sampling_rate=16000, return_tensors="pt", padding=True)
-    with torch.no_grad():
-        logits = model(**inputs).logits
-    predicted_ids = torch.argmax(logits, dim=-1)
-    transcription = processor.batch_decode(predicted_ids)[0]
-    return transcription.lower()
+@app.post("/transcribe")
+async def transcribe_audio(request: Request):
+    try:
+        data = await request.json()
+        audio_path = None
+
+        if "base64" in data:
+            audio_bytes = base64.b64decode(data["base64"])
+            with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as tmp:
+                tmp.write(audio_bytes)
+                audio_path = tmp.name
+
+        elif "url" in data:
+            response = requests.get(data["url"])
+            with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as tmp:
+                tmp.write(response.content)
+                audio_path = tmp.name
+        else:
+            return JSONResponse({"error": "Debés enviar 'base64' o 'url'"}, status_code=400)
 
-# Interfaz de Gradio
-demo = gr.Interface(
-    fn=transcribir,
-    inputs=gr.Audio(type="filepath"),
-    outputs="text",
-    title="Transcriptor Guaraní",
-    description="Subí un audio en guaraní (.ogg, .wav) y obtené la transcripción"
-)
+        # Transcripción
+        audio_data, _ = librosa.load(audio_path, sr=16000)
+        inputs = processor(audio_data, sampling_rate=16000, return_tensors="pt", padding=True)
+        with torch.no_grad():
+            logits = model(**inputs).logits
+        predicted_ids = torch.argmax(logits, dim=-1)
+        transcription = processor.batch_decode(predicted_ids)[0]
 
-demo.launch()
+        return JSONResponse({"text": transcription.lower()})
+    except Exception as e:
+        return JSONResponse({"error": str(e)}, status_code=500)