app.py

import gradio as gr
from PIL import Image
import os

#from diffusers import StableDiffusionPipeline
whisper = gr.Interface.load(name="spaces/sanchit-gandhi/whisper-large-v2")
#stable_diffusion = gr.Blocks.load(name="spaces/stabilityai/stable-diffusion")
stable_diffusion = gr.Blocks.load(name="spaces/runwayml/stable-diffusion-v1-5")
### ————————————————————————————————————————
title="Talking to Stable Diffusion"
### ————————————————————————————————————————
def get_images(prompt):
    #gallery_dir = stable_diffusion(prompt, None, None, fn_index=2)
    gallery_dir = stable_diffusion(prompt, fn_index=2)
    return [os.path.join(gallery_dir, img) for img in os.listdir(gallery_dir)]


def translate_better(audio):
    print("""
    —
    Sending audio to Whisper ...
    —
    """)
    transcribe_text_result = whisper(audio, None, "transcribe", fn_index=0)
    translate_text_result = whisper(audio, None, "translate", fn_index=0)
    print("transcript: " + transcribe_text_result)
    print("———————————————————————————————————————————")  
    print("translated: " + translate_text_result)

    return transcribe_text_result, translate_text_result


 
with gr.Blocks() as demo:
    gr.Markdown(
            """
            ## 1. Say what you want:
            """
        )
    with gr.Column():
        with gr.Tab(label="Record audio input", elem_id="record_tab"):
            with gr.Column():
                record_input = gr.Audio(
                                    source="microphone",
                                    type="filepath", 
                                    show_label=False,
                                    elem_id="record_btn"
                                )
                with gr.Row():
                    audio_r_translate = gr.Button("Check Whisper first", elem_id="check_btn_1")              
                    audio_r_direct_sd = gr.Button("Generating Images", elem_id="magic_btn_1")
        
        with gr.Accordion(label="Stable Diffusion Settings", elem_id="sd_settings", visible=False):
            with gr.Row():
                guidance_scale = gr.Slider(2, 15, value = 7, label = 'Guidance Scale')
                nb_iterations = gr.Slider(10, 50, value = 25, step = 1, label = 'Steps')
                seed = gr.Slider(label = "Seed", minimum = 0, maximum = 2147483647, step = 1, randomize = True)
        
        gr.Markdown(
            """
            ## 2. Check Whisper output:
            """
        )
        
        with gr.Row():
            transcripted_output = gr.Textbox(
                                    label="Transcription in your detected spoken language", 
                                    lines=3,
                                    elem_id="transcripted"
                                )
            translated_output = gr.Textbox(
                                    label="Transcription in your detected spoken language", 
                                    lines=3,
                                    elem_id="translated"
                                )
           
            
                    
        gr.Markdown("""
            ## 3. Wait for Stable Diffusion Results about ~10 seconds
            """
            ) 
        
        sd_output = gr.Gallery().style(grid=2, height="auto")
        audio_r_translate.click(translate_better, 
                                inputs = [
                                    record_input
                                ], 
                                outputs = [
                                    transcripted_output,
                                    translated_output,
                                ])
        audio_r_direct_sd.click(get_images, 
                              inputs = [
                                  translated_output
                                  ], 
                              outputs = sd_output
                          )
    
if __name__ == "__main__":
    demo.queue(max_size=32, concurrency_count=20).launch()