First draft

app.py

@@ -1,13 +1,20 @@
 import gradio as gr
 import spaces
 import torch
+from transformers.pipelines.audio_utils import ffmpeg_read
+import io
+import whisper
 
-zero = torch.Tensor([0]).cuda()
-print(zero.device) # <-- 'cpu' 🤔
+model = whisper.load_model("large", device="cuda")
+writer = whisper.utils.get_writer("srt", "/dev/null")
 
 @spaces.GPU
-def greet(n):
-    print(zero.device) # <-- 'cuda:0' 🤗
-    return f"Hello {zero + n} Tensor"
+def generate(file):
+    # get file to type bytes somehow
+    audio = ffmpeg_read(file)
+    result = model.transcribe(audio)
+    out = io.StringIO()
+    writer.write_result(result, out)
+    return out.getvalue()
 
-gr.Interface(fn=greet, inputs=gr.Number(), outputs=gr.Text()).launch()
+gr.Interface(fn=greet, inputs=gr.File(type="binary"), outputs=gr.Text()).launch()