ASR for Local Languages

app.py

@@ -0,0 +1,90 @@
+import traceback
+import soundfile as sf
+import torch
+import numpy as np
+from transformers import AutoProcessor, AutoModelForSpeechSeq2Seq
+import gradio as gr
+import resampy
+
+# Language code mapping
+LANGUAGE_CODES = {
+    "Amharic": "amh",
+    "Swahili": "swh",
+    "Somali": "som",
+    "Afan Oromo": "orm",
+    "Tigrinya": "tir",
+    "Chichewa": "nya"
+}
+
+# --- Load ASR model ---
+try:
+    model_id = "facebook/seamless-m4t-v2-large"
+    processor = AutoProcessor.from_pretrained(model_id)
+    asr_model = AutoModelForSpeechSeq2Seq.from_pretrained(model_id).to("cpu")
+    print("[INFO] ASR model loaded successfully.")
+except Exception as e:
+    print("[ERROR] Failed to load ASR model:", e)
+    traceback.print_exc()
+    asr_model = None
+    processor = None
+
+# --- Helper: ASR ---
+def transcribe_audio(audio_file, language):
+    if asr_model is None or processor is None:
+        return "ASR Model loading failed"
+    
+    try:
+        # Get language code
+        lang_code = LANGUAGE_CODES.get(language)
+        if not lang_code:
+            return f"Unsupported language: {language}"
+        
+        # Read and preprocess audio
+        audio, sr = sf.read(audio_file)
+        if audio.ndim > 1:
+            audio = audio.mean(axis=1)
+        audio = resampy.resample(audio, sr, 16000)
+        
+        # Process with model
+        inputs = processor(audios=audio, sampling_rate=16000, return_tensors="pt")
+        
+        with torch.no_grad():
+            generated_ids = asr_model.generate(**inputs, tgt_lang=lang_code)
+        
+        # Decode the transcription
+        transcription = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]
+        return transcription.strip()
+    
+    except Exception as e:
+        print(f"[ERROR] ASR transcription failed for {language}:", e)
+        traceback.print_exc()
+        return f"ASR failed: {str(e)[:50]}..."
+
+# --- Gradio UI ---
+with gr.Blocks(title="🌍 Multilingual ASR") as demo:
+    gr.Markdown("# 🌍 Multilingual Speech Recognition")
+    gr.Markdown("Transcribe audio in Amharic, Swahili, Somali, Afan Oromo, Tigrinya, or Chichewa")
+    
+    with gr.Row():
+        with gr.Column():
+            audio_input = gr.Audio(sources=["microphone", "upload"], type="filepath", label="Record or upload audio")
+            language_select = gr.Dropdown(
+                choices=list(LANGUAGE_CODES.keys()),
+                value="Swahili",
+                label="Select Language"
+            )
+    
+    submit_btn = gr.Button("Transcribe", variant="primary")
+    
+    with gr.Row():
+        with gr.Column():
+            transcription_output = gr.Textbox(label="Transcription")
+    
+    submit_btn.click(
+        fn=transcribe_audio,
+        inputs=[audio_input, language_select],
+        outputs=transcription_output
+    )
+
+if __name__ == "__main__":
+    demo.launch(server_name="0.0.0.0", server_port=7860)

requirements.txt

@@ -0,0 +1,12 @@
+torch
+torchaudio
+transformers
+gradio
+soundfile
+resampy
+accelerate
+sentencepiece
+scipy
+numpy
+sacremoses
+librosa