update3

app.py

@@ -1,37 +1,54 @@
-# app.py – NÚNA VIRKAR Á ZeroGPU (A100/T4)
+# app.py – Virkar á ZeroGPU – allt að 3 mínútur (180 sek)
 import os
 import gradio as gr
-import spaces  # MÁGÍSKUR – virkjar ZeroGPU
+import spaces
 from transformers import pipeline
+import numpy as np
 
-# ÞITT private model
 MODEL_NAME = "palli23/whisper-small-sam_spjall"
 
-# Þetta fall keyrir á GPU í ZeroGPU
 @spaces.GPU
-def transcribe(audio):
-    if not audio:
+def transcribe_long(audio_path):
+    if not audio_path:
         return "Hladdu upp hljóðskrá"
     
-    # Módel hleðst á GPU í fyrsta skipti (cached)
+    # Hlaða ASR á GPU (cached)
     pipe = pipeline(
         "automatic-speech-recognition",
         model=MODEL_NAME,
-        device=0,                    # 0 = GPU (ZeroGPU gefur þér A100/T4)
+        device=0,
         token=os.getenv("HF_TOKEN")
     )
-    result = pipe(audio, batch_size=8)
+    
+    # Chunkar – 30 sek + 2 sek overlap (besta niðurstaðan fyrir Whisper-small)
+    chunk_length_s = 30
+    stride_length_s = 2
+    
+    result = pipe(
+        audio_path,
+        chunk_length_s=chunk_length_s,
+        stride_length_s=(stride_length_s, stride_length_s),
+        return_timestamps=False,
+        batch_size=8
+    )
+    
     return result["text"]
 
-# Einfalt interface
-with gr.Blocks(title="Íslenskt ASR – ZeroGPU Live") as demo:
+# Gradio – fallegt og tilbúið fyrir 3 mín klippur
+with gr.Blocks(title="Íslenskt ASR – 3 mínútur") as demo:
     gr.Markdown("# Íslenskt ASR – ZeroGPU (A100/T4)")
-    gr.Markdown("**~4 % WER · 0.2× RTF · Keyrir á GPU núna**")
+    gr.Markdown("**Styður allt að 3 mínútna hljóðklippur · ~4 % WER · 25–45 sek transcribe**")
+    
+    audio = gr.Audio(
+        type="filepath",
+        label="Hladdu upp .mp3 / .wav / .m4a (allt að 3 mín)"
+    )
+    
+    btn = gr.Button("Transcribe (25–45 sek)", variant="primary", size="lg")
+    out = gr.Textbox(lines=30, label="Útskrift")
     
-    audio = gr.Audio(type="filepath", label="Hladdu upp .mp3 / .wav")
-    btn = gr.Button("Transcribe (15–25 sek)", variant="primary", size="lg")
-    out = gr.Textbox(lines=25, label="Útskrift")
+    btn.click(transcribe_long, inputs=audio, outputs=out)
     
-    btn.click(transcribe, audio, out)
+    gr.Markdown("© 2025 – Einkaeign")
 
 demo.launch(auth=("beta", "beta2025"))