app.py

from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
import torch
import librosa
import gradio as gr

# Cargamos el modelo de 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()

# 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"
)