app.py

import gradio as gr

import torch

from spectro import wav_bytes_from_spectrogram_image
from diffusers import StableDiffusionPipeline

from transformers import BlipForConditionalGeneration, BlipProcessor

from share_btn import community_icon_html, loading_icon_html, share_js

model_id = "riffusion/riffusion-model-v1"
blip_model_id = "Salesforce/blip-image-captioning-base"
pipe = StableDiffusionPipeline.from_pretrained(model_id)
pipe = pipe.to("cuda")

blip_model = BlipForConditionalGeneration.from_pretrained(blip_model_id, torch_dtype=torch.float16).to("cuda")
processor = BlipProcessor.from_pretrained(blip_model_id)

def predict(image):
    inputs = processor(image, return_tensors="pt").to("cuda", torch.float16)
    output_blip = blip_model.generate(**inputs)
    prompt = processor.decode(output_blip[0], skip_special_tokens=True)
    
    spec = pipe(prompt).images[0]
    print(spec)
    wav = wav_bytes_from_spectrogram_image(spec)
    with open("output.wav", "wb") as f:
        f.write(wav[0].getbuffer())
    return spec, 'output.wav', gr.update(visible=True), gr.update(visible=True), gr.update(visible=True)

title = """
    <div style="text-align: center; max-width: 500px; margin: 0 auto;">
        <div
        style="
            display: inline-flex;
            align-items: center;
            gap: 0.8rem;
            font-size: 1.75rem;
            margin-bottom: 10px;
        "
        >
        <h1 style="font-weight: 600; margin-bottom: 7px;">
            Riffusion real-time image-to-music generation
        </h1>
        </div>
        <p style="margin-bottom: 10px;font-size: 94%;font-weight: 100;line-height: 1.5em;">
        Describe a musical prompt, generate music by getting a spectrogram image & sound.
    </div>
"""

article = """
    <p style="font-size: 0.8em;line-height: 1.2em;border: 1px solid #374151;border-radius: 8px;padding: 20px;">
    About the model: Riffusion is a latent text-to-image diffusion model capable of generating spectrogram images given any text input. These spectrograms can be converted into audio clips.
    <br />—
    <br />The Riffusion model was created by fine-tuning the Stable-Diffusion-v1-5 checkpoint.
    <br />—
    <br />The model is intended for research purposes only. Possible research areas and tasks include 
    generation of artworks, audio, and use in creative processes, applications in educational or creative tools, research on generative models.

    </p>
    <div class="footer">
        <p>
        <a href="https://huggingface.co/riffusion/riffusion-model-v1" target="_blank">Riffusion model</a> by Seth Forsgren and Hayk Martiros - 
        <a href="https://github.com/salesforce/BLIP" target="_blank"> BLIP Model </a> by Junnan Li et al. - Demo forked from 🤗 <a href="https://twitter.com/fffiloni" target="_blank">Sylvain Filoni</a>'s demo
        </p>
    </div>

    <p style="text-align: center;font-size: 94%">
        Do you need faster results ? You can skip the queue by duplicating this space: 
        <span style="display: flex;align-items: center;justify-content: center;height: 30px;">
        <a href="https://huggingface.co/fffiloni/spectrogram-to-music?duplicate=true"><img src="https://img.shields.io/badge/-Duplicate%20Space-blue?labelColor=white&style=flat&logo=data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAABAAAAAQCAYAAAAf8/9hAAAAAXNSR0IArs4c6QAAAP5JREFUOE+lk7FqAkEURY+ltunEgFXS2sZGIbXfEPdLlnxJyDdYB62sbbUKpLbVNhyYFzbrrA74YJlh9r079973psed0cvUD4A+4HoCjsA85X0Dfn/RBLBgBDxnQPfAEJgBY+A9gALA4tcbamSzS4xq4FOQAJgCDwV2CPKV8tZAJcAjMMkUe1vX+U+SMhfAJEHasQIWmXNN3abzDwHUrgcRGmYcgKe0bxrblHEB4E/pndMazNpSZGcsZdBlYJcEL9Afo75molJyM2FxmPgmgPqlWNLGfwZGG6UiyEvLzHYDmoPkDDiNm9JR9uboiONcBXrpY1qmgs21x1QwyZcpvxt9NS09PlsPAAAAAElFTkSuQmCC&logoWidth=14" alt="Duplicate Space"></a>       
        </span>
    </p>
"""

css = '''
    #col-container, #col-container-2 {max-width: 510px; margin-left: auto; margin-right: auto;}
    a {text-decoration-line: underline; font-weight: 600;}
    div#record_btn > .mt-6 {
        margin-top: 0!important;
    }
    div#record_btn > .mt-6 button {
        width: 100%;
        height: 40px;
    }
    .footer {
        margin-bottom: 45px;
        margin-top: 10px;
        text-align: center;
        border-bottom: 1px solid #e5e5e5;
    }
    .footer>p {
        font-size: .8rem;
        display: inline-block;
        padding: 0 10px;
        transform: translateY(10px);
        background: white;
    }
    .dark .footer {
        border-color: #303030;
    }
    .dark .footer>p {
        background: #0b0f19;
    }
    .animate-spin {
        animation: spin 1s linear infinite;
    }
    @keyframes spin {
        from {
            transform: rotate(0deg);
        }
        to {
            transform: rotate(360deg);
        }
    }
    #share-btn-container {
        display: flex; padding-left: 0.5rem !important; padding-right: 0.5rem !important; background-color: #000000; justify-content: center; align-items: center; border-radius: 9999px !important; width: 13rem;
    }
    #share-btn {
        all: initial; color: #ffffff;font-weight: 600; cursor:pointer; font-family: 'IBM Plex Sans', sans-serif; margin-left: 0.5rem !important; padding-top: 0.25rem !important; padding-bottom: 0.25rem !important;right:0;
    }
    #share-btn * {
        all: unset;
    }
    #share-btn-container div:nth-child(-n+2){
        width: auto !important;
        min-height: 0px !important;
    }
    #share-btn-container .wrap {
        display: none !important;
    }

'''
 


with gr.Blocks(css=css) as demo:
    
    with gr.Column(elem_id="col-container"):
        
        gr.HTML(title)
        
        # prompt_input = gr.Textbox(placeholder="a cat diva singing in a New York jazz club", label="Musical prompt", elem_id="prompt-in")
        image_input = gr.Image()
        send_btn = gr.Button(value="Get a new spectrogram ! ", elem_id="submit-btn")
            
    with gr.Column(elem_id="col-container-2"):
        
        spectrogram_output = gr.Image(label="spectrogram image result", elem_id="img-out")
        sound_output = gr.Audio(type='filepath', label="spectrogram sound", elem_id="music-out")
        
        with gr.Group(elem_id="share-btn-container"):
            community_icon = gr.HTML(community_icon_html, visible=False)
            loading_icon = gr.HTML(loading_icon_html, visible=False)
            share_button = gr.Button("Share to community", elem_id="share-btn", visible=False)
        
        gr.HTML(article)
    
    send_btn.click(predict, inputs=[image_input], outputs=[spectrogram_output, sound_output, share_button, community_icon, loading_icon])
    share_button.click(None, [], [], _js=share_js)

demo.queue(max_size=250).launch(debug=True)