Update app.py

app.py

@@ -105,7 +105,9 @@ def transcribe_audio(audio_file, chunk_length_s=30):
             chunk_waveform = torch.mean(chunk_waveform, dim=0, keepdim=True)
 
         # Process the chunk with the tokenizer
+        processor = tokenizer.add_special_tokens({'pad_token': '[PAD]'})
         inputs = processor(chunk_waveform.squeeze(0).numpy(), sampling_rate=sample_rate, return_tensors="pt")
+        
         input_features = inputs.input_features
 
         # Create attention mask
@@ -257,7 +259,7 @@ def save_to_pdf(text, summary):
 iface = gr.Blocks()
 
 with iface:
-    gr.HTML('<img src="https://huggingface.co/spaces/camparchimedes/ola_s-audioshop/resolve/main/banner_trans.png" width="100%" height="auto"/>')
+    gr.HTML('<img src="https://huggingface.co/spaces/camparchimedes/ola_s-audioshop/raw/main/banner_trans.png" width="100%" height="auto"/>')
     gr.Markdown("**Switch Work's JoJo-versjon som webapp for transkribering av lydfiler til norsk skrift. Språkmodell: NbAiLab/nb-whisper-large, Ekstra: oppsummering, pdf-download**")
 
     with gr.Tabs():