Another go

app.py

@@ -15,14 +15,15 @@ YT_LENGTH_LIMIT_S = 3600  # limit to 1 hour YouTube files
 from transformers import AutoTokenizer
 
 tokenizer = AutoTokenizer.from_pretrained("openai/whisper-large-v3")
-assert tokenizer.is_fast
-tokenizer.save_pretrained("...")
+# assert tokenizer.is_fast
+# tokenizer.save_pretrained("...")
 
 
 device = 0 if torch.cuda.is_available() else "cpu"
 
 pipe = pipeline(
     task="automatic-speech-recognition",
+    tokenizer=tokenizer,
     model=MODEL_NAME,
     chunk_length_s=30,
     device=device,