Update app.py

app.py

@@ -1,17 +1,35 @@
+import torch
+from transformers import AutoModelForSpeechSeq2Seq, AutoProcessor, pipeline
 import gradio as gr
-from transformers import pipeline
 
-# Use whisper-base for better Swedish transcription on CPU
-asr = pipeline("automatic-speech-recognition", model="openai/whisper-base", device=-1)
+device = "cuda" if torch.cuda.is_available() else "cpu"
+torch_dtype = torch.float16 if torch.cuda.is_available() else torch.float32
+
+model_id = "KBLab/kb-whisper-large"
+
+model = AutoModelForSpeechSeq2Seq.from_pretrained(
+    model_id, torch_dtype=torch_dtype
+).to(device)
+
+processor = AutoProcessor.from_pretrained(model_id)
+
+pipe = pipeline(
+    "automatic-speech-recognition",
+    model=model,
+    tokenizer=processor.tokenizer,
+    feature_extractor=processor.feature_extractor,
+    device=device,
+    torch_dtype=torch_dtype,
+)
 
 def transcribe(audio):
-    result = asr(audio)
+    result = pipe(audio, chunk_length_s=30, generate_kwargs={"task": "transcribe", "language": "sv"})
     return result["text"]
 
 gr.Interface(
     fn=transcribe,
     inputs=gr.Audio(source="upload", type="filepath", label="Upload Swedish Audio"),
-    outputs=gr.Textbox(label="Transcribed Text (Swedish)"),
-    title="Swedish Whisper Transcriber",
-    description="This app uses OpenAI Whisper (base) to transcribe spoken Swedish into text for free."
+    outputs=gr.Textbox(label="Transcribed Text"),
+    title="KB-Whisper Swedish Transcriber",
+    description="Transcribes Swedish audio using KBLab's Whisper Large model."
 ).launch()