File size: 8,405 Bytes
929aad9
 
 
 
 
 
 
c879843
cda11a5
 
 
 
 
 
 
929aad9
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
c879843
 
 
 
 
 
 
929aad9
a3fc4ad
929aad9
cda11a5
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
929aad9
 
 
 
 
 
 
 
 
 
 
 
cda11a5
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
929aad9
a8c3afa
 
 
 
 
 
 
 
929aad9
cda11a5
 
 
 
 
 
929aad9
 
 
a8c3afa
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
import torch
import torchaudio
from einops import rearrange
import gradio as gr
import spaces
import os
import uuid
from pydub import AudioSegment
import numpy as np
import random
import torch
from diffusers import StableDiffusion3Pipeline, SD3Transformer2DModel, FlowMatchEulerDiscreteScheduler


'''AUDIO'''

# Importing the model-related functions
from stable_audio_tools import get_pretrained_model
from stable_audio_tools.inference.generation import generate_diffusion_cond

# Load the model outside of the GPU-decorated function
def load_model():
    print("Loading model...")
    model, model_config = get_pretrained_model("stabilityai/stable-audio-open-1.0")
    print("Model loaded successfully.")
    return model, model_config

# Function to set up, generate, and process the audio
@spaces.GPU(duration=120)  # Allocate GPU only when this function is called
def generate_audio(prompt, seconds_total=30, steps=100, cfg_scale=7):
    print(f"Prompt received: {prompt}")
    print(f"Settings: Duration={seconds_total}s, Steps={steps}, CFG Scale={cfg_scale}")

    device = "cuda" if torch.cuda.is_available() else "cpu"
    print(f"Using device: {device}")

    # Fetch the Hugging Face token from the environment variable
    hf_token = os.getenv('HF_TOKEN')
    print(f"Hugging Face token: {hf_token}")

    # Use pre-loaded model and configuration
    model, model_config = load_model()
    sample_rate = model_config["sample_rate"]
    sample_size = model_config["sample_size"]

    print(f"Sample rate: {sample_rate}, Sample size: {sample_size}")

    model = model.to(device)
    print("Model moved to device.")

    # Set up text and timing conditioning
    conditioning = [{
        "prompt": prompt,
        "seconds_start": 0,
        "seconds_total": seconds_total
    }]
    print(f"Conditioning: {conditioning}")

    # Generate stereo audio
    print("Generating audio...")
    output = generate_diffusion_cond(
        model,
        steps=steps,
        cfg_scale=cfg_scale,
        conditioning=conditioning,
        sample_size=sample_size,
        sigma_min=0.3,
        sigma_max=500,
        sampler_type="dpmpp-3m-sde",
        device=device
    )
    print("Audio generated.")

    # Rearrange audio batch to a single sequence
    output = rearrange(output, "b d n -> d (b n)")
    print("Audio rearranged.")

    # Peak normalize, clip, convert to int16
    output = output.to(torch.float32).div(torch.max(torch.abs(output))).clamp(-1, 1).mul(32767).to(torch.int16).cpu()
    print("Audio normalized and converted.")

    # Generate a unique filename for the output
    unique_filename = f"output_{uuid.uuid4().hex}.wav"
    print(f"Saving audio to file: {unique_filename}")

    # Save to file
    torchaudio.save(unique_filename, output, sample_rate)
    print(f"Audio saved: {unique_filename}")

    # Convert WAV to MP3 using pydub without ffmpeg
    audio = AudioSegment.from_wav(unique_filename)
    full_path_mp3 = unique_filename.replace('wav', 'mp3')
    audio.export(full_path_mp3, format="mp3")
    
    print(f"Audio converted and saved to MP3: {full_path_mp3}")

    # Return the path to the generated audio file
    return full_path_mp3

'''DIFFUSION'''
device = "cuda" if torch.cuda.is_available() else "cpu"
dtype = torch.float16

repo = "stabilityai/stable-diffusion-3-medium-diffusers"
pipe = StableDiffusion3Pipeline.from_pretrained(repo, torch_dtype=torch.float16).to(device)

MAX_SEED = np.iinfo(np.int32).max
MAX_IMAGE_SIZE = 1344

@spaces.GPU
def infer(prompt, negative_prompt, seed, randomize_seed, width, height, guidance_scale, num_inference_steps, progress=gr.Progress(track_tqdm=True)):

    if randomize_seed:
        seed = random.randint(0, MAX_SEED)
        
    generator = torch.Generator().manual_seed(seed)
    
    image = pipe(
        prompt = prompt, 
        negative_prompt = negative_prompt,
        guidance_scale = guidance_scale, 
        num_inference_steps = num_inference_steps, 
        width = width, 
        height = height,
        generator = generator
    ).images[0] 
    
    return image, seed

'''
# Setting up the Gradio Interface
interface = gr.Interface(
    fn=generate_audio,
    inputs=[
        gr.Textbox(label="Prompt", placeholder="Enter your text prompt here"),
        gr.Slider(0, 47, value=30, label="Duration in Seconds"),
        gr.Slider(10, 150, value=100, step=10, label="Number of Diffusion Steps"),
        gr.Slider(1, 15, value=7, step=0.1, label="CFG Scale")
    ],
    outputs=gr.Audio(type="filepath", label="Generated Audio"),
    title="Stable Audio Generator",
    description="Generate variable-length stereo audio at 44.1kHz from text prompts using Stable Audio Open 1.0.",
)'''

with gr.Blocks() as demo:
    with gr.Tab("SD3"):
        with gr.Column:
            gr.Markdown(f"""
            # Demo [Stable Diffusion 3 Medium](https://huggingface.co/stabilityai/stable-diffusion-3-medium)
            Learn more about the [Stable Diffusion 3 series](https://stability.ai/news/stable-diffusion-3). Try on [Stability AI API](https://platform.stability.ai/docs/api-reference#tag/Generate/paths/~1v2beta~1stable-image~1generate~1sd3/post), [Stable Assistant](https://stability.ai/stable-assistant), or on Discord via [Stable Artisan](https://stability.ai/stable-artisan). Run locally with [ComfyUI](https://github.com/comfyanonymous/ComfyUI) or [diffusers](https://github.com/huggingface/diffusers)
            """)
        with gr.Row():
            prompt = gr.Text(
                label="Prompt",
                show_label=False,
                max_lines=1,
                placeholder="Enter your prompt",
                container=False,
            )
            run_button = gr.Button("Run", scale=0)
        result = gr.Image(label="Result", show_label=False)

        with gr.Accordion("Advanced Settings", open=False):
            
            negative_prompt = gr.Text(
                label="Negative prompt",
                max_lines=1,
                placeholder="Enter a negative prompt",
            )
            
            seed = gr.Slider(
                label="Seed",
                minimum=0,
                maximum=MAX_SEED,
                step=1,
                value=0,
            )
            
            randomize_seed = gr.Checkbox(label="Randomize seed", value=True)
            
            with gr.Row():
                
                width = gr.Slider(
                    label="Width",
                    minimum=256,
                    maximum=MAX_IMAGE_SIZE,
                    step=64,
                    value=1024,
                )
                
                height = gr.Slider(
                    label="Height",
                    minimum=256,
                    maximum=MAX_IMAGE_SIZE,
                    step=64,
                    value=1024,
                )
            
            with gr.Row():
                
                guidance_scale = gr.Slider(
                    label="Guidance scale",
                    minimum=0.0,
                    maximum=10.0,
                    step=0.1,
                    value=5.0,
                )
                
                num_inference_steps = gr.Slider(
                    label="Number of inference steps",
                    minimum=1,
                    maximum=50,
                    step=1,
                    value=28,
                )

    with gr.Tab("Audio"):
        audio_prompt =  gr.Textbox(label="Prompt", placeholder="Enter your text prompt here")
        audio_duration = gr.Slider(0, 47, value=30, label="Duration in Seconds")
        audio_steps =  gr.Slider(10, 150, value=100, step=10, label="Number of Diffusion Steps")
        audio_cfg = gr.Slider(1, 15, value=7, step=0.1, label="CFG Scale")
        audio_process_button = gr.Button("Process Audio")
        audio_output = gr.Audio(type="filepath", label="Generated Audio")
    audio_process_button.click(generate_audio, [audio_prompt, audio_duration, audio_steps, audio_cfg], [audio_output])

    gr.on(
        triggers=[run_button.click, prompt.submit, negative_prompt.submit],
        fn = infer,
        inputs = [prompt, negative_prompt, seed, randomize_seed, width, height, guidance_scale, num_inference_steps],
        outputs = [result, seed]
    )
# Pre-load the model to avoid multiprocessing issues
model, model_config = load_model()

demo.launch()