Upload 3 files

app.py

@@ -1,10 +1,10 @@
 import gradio as gr
 import torch
+import whisper
 from diffusers import DiffusionPipeline
 from transformers import (
     WhisperForConditionalGeneration,
     WhisperProcessor,
-    pipeline,
 )
 
 import os
@@ -14,12 +14,10 @@ device = "cuda" if torch.cuda.is_available() else "cpu"
 model = WhisperForConditionalGeneration.from_pretrained("whispy/whisper_italian").to(device)
 processor = WhisperProcessor.from_pretrained("whispy/whisper_italian")
 
-pipe = pipeline(model="whispy/whisper_italian")
-
 diffuser_pipeline = DiffusionPipeline.from_pretrained(
     "CompVis/stable-diffusion-v1-4",
     custom_pipeline="speech_to_image_diffusion",
-    speech_model="whispy/whisper_italian",
+    speech_model=model,
     speech_processor=processor,
     use_auth_token=MY_SECRET_TOKEN,
     revision="fp16",
@@ -29,10 +27,6 @@ diffuser_pipeline = DiffusionPipeline.from_pretrained(
 diffuser_pipeline.enable_attention_slicing()
 diffuser_pipeline = diffuser_pipeline.to(device)
 
-def transcribe(audio):
-    text = pipe(audio)["text"]
-    return text
-
 
 #————————————————————————————————————————————
 # GRADIO SETUP
@@ -51,8 +45,7 @@ image_output = gr.Image()
 
 def speech_to_text(audio_sample):
   
-  #process_audio = whisper.load_audio(audio_sample)
-  process_audio = transcribe(audio_sample)
+  process_audio = whisper.load_audio(audio_sample)
   output = diffuser_pipeline(process_audio)
  
   print(f"""

requirements.txt

@@ -1,7 +1,7 @@
-transformers
-torch
 --extra-index-url https://download.pytorch.org/whl/cu113
 torch
 scipy
 ftfy
-diffusers
+git+https://github.com/huggingface/transformers
+git+https://github.com/huggingface/diffusers
+git+https://github.com/openai/whisper.git