create app.py

app.py

@@ -0,0 +1,82 @@
+# app.py
+
+import gradio as gr
+from transformers import pipeline
+import soundfile as sf
+import numpy as np
+
+# --- EDIT THIS: map display names to your HF Hub model IDs ---
+language_models = {
+    "Akan (Asanti Twi)": "FarmerlineML/w2v-bert-2.0_twi_alpha_v1",
+    "Akan (Fante Twi)": "your-username/english-asr-model",
+    "Ewe":  "FarmerlineML/w2v-bert-2.0_ewe_2",
+    "Kiswahili": "FarmerlineML/w2v-bert-2.0_swahili_alpha",
+    "Luganda": "FarmerlineML/w2v-bert-2.0_luganda",
+    "Akan (Asanti Twi)": "your-username/english-asr-model",
+    "Brazilian Portuguese": "FarmerlineML/w2v-bert-2.0_brazilian_portugese_alpha",
+    # add more as needed
+}
+
+# Pre-load pipelines for each language
+asr_pipelines = {
+    lang: pipeline(
+        task="automatic-speech-recognition",
+        model=model_id,
+        # device=0,           # uncomment if you have GPU
+        chunk_length_s=30      # adjust if your audio can be longer
+    )
+    for lang, model_id in language_models.items()
+}
+
+
+def transcribe(audio_path: str, language: str) -> str:
+    """
+    Load the audio file, convert to mono if needed,
+    and run it through the selected ASR pipeline.
+    """
+    if audio_path is None:
+        return "⚠️ Please upload or record an audio clip."
+
+    # Read the file
+    speech, sr = sf.read(audio_path)
+    # Stereo → mono
+    if speech.ndim > 1:
+        speech = np.mean(speech, axis=1)
+
+    result = asr_pipelines[language]({
+        "sampling_rate": sr,
+        "raw": speech
+    })
+    return result.get("text", "")
+
+
+with gr.Blocks(title="🌐 Multilingual ASR Demo") as demo:
+    gr.Markdown(
+        """
+        ## 🎙️ Multilingual Speech-to-Text   
+        Upload an audio file or record via your microphone,  
+        then choose the language/model and hit **Transcribe**.
+        """
+    )
+
+    with gr.Row():
+        lang = gr.Dropdown(
+            choices=list(language_models.keys()),
+            value=list(language_models.keys())[0],
+            label="Select Language / Model"
+        )
+
+    with gr.Row():
+        audio = gr.Audio(
+            source=["upload", "microphone"],
+            type="filepath",
+            label="Upload or Record Audio"
+        )
+
+    btn = gr.Button("Transcribe")
+    output = gr.Textbox(label="Transcription")
+
+    btn.click(fn=transcribe, inputs=[audio, lang], outputs=output)
+
+if __name__ == "__main__":
+    demo.launch()