File size: 18,862 Bytes
9fd51b2
3bddfce
d50061a
 
9fd51b2
d50061a
9fd51b2
3bddfce
9fd51b2
3bddfce
9fd51b2
d50061a
3a7d7b2
4b4ce6b
 
 
 
 
 
d50061a
9fd51b2
3bddfce
9fd51b2
3bddfce
 
9fd51b2
4b4ce6b
 
3bddfce
 
 
 
 
4b4ce6b
9fd51b2
d50061a
 
3bddfce
9fd51b2
e4ac605
9fd51b2
3bddfce
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4b4ce6b
d50061a
7a2d549
 
 
 
 
d50061a
 
 
23ffa3b
d50061a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4b4ce6b
 
 
 
 
d50061a
4b4ce6b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3b6f2b0
 
 
 
23e2bc5
 
 
 
4b4ce6b
 
 
 
 
 
 
 
 
 
 
 
 
7a2d549
 
 
 
 
4b4ce6b
9bf1244
7236a22
9bf1244
 
 
862f243
9bf1244
 
7236a22
9bf1244
 
 
862f243
 
 
 
 
 
 
 
 
 
 
 
7236a22
862f243
 
 
 
 
 
 
 
 
 
 
 
 
9bf1244
4b4ce6b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
7236a22
9bf1244
23ffa3b
9bf1244
 
 
862f243
4b4ce6b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3bddfce
 
 
 
 
 
 
 
4b4ce6b
3bddfce
 
 
 
 
 
 
 
4b4ce6b
3bddfce
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4b4ce6b
3bddfce
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4b4ce6b
3bddfce
 
 
 
 
d50061a
4b4ce6b
3bddfce
 
 
 
 
 
4b4ce6b
3bddfce
 
 
 
 
 
4b4ce6b
 
9bf1244
3bddfce
 
 
 
 
 
 
 
 
4b4ce6b
3bddfce
4b4ce6b
 
3bddfce
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4b4ce6b
3bddfce
 
 
 
 
 
 
 
 
 
4b4ce6b
3bddfce
 
d50061a
4b4ce6b
 
d50061a
4b4ce6b
 
 
 
d50061a
4b4ce6b
 
d50061a
4b4ce6b
 
 
 
 
 
 
 
 
 
 
 
d50061a
4b4ce6b
 
 
 
 
 
 
 
 
 
 
 
 
d50061a
4b4ce6b
 
 
 
 
3bddfce
4b4ce6b
3bddfce
 
4b4ce6b
 
3bddfce
 
 
 
 
 
4b4ce6b
 
 
438b9a5
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
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
import gradio as gr
#import torch
#import whisper
#from datetime import datetime
from PIL import Image
#import flag
import os
#MY_SECRET_TOKEN=os.environ.get('HF_TOKEN_SD')

#from diffusers import StableDiffusionPipeline

whisper = gr.Interface.load(name="spaces/sanchit-gandhi/whisper-large-v2")
stable_diffusion = gr.Blocks.load(name="spaces/runwayml/stable-diffusion-v1-5")
### β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”

title="Whisper to Stable Diffusion"

### β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”

#whisper_model = whisper.load_model("small")

#device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")

#pipe = StableDiffusionPipeline.from_pretrained("CompVis/stable-diffusion-v1-4", use_auth_token=MY_SECRET_TOKEN)
#pipe.to(device)

### β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”

def get_images(prompt):
    gallery_dir = stable_diffusion(prompt, fn_index=2)
    return [os.path.join(gallery_dir, img) for img in os.listdir(gallery_dir)]


def magic_whisper_to_sd(audio, guidance_scale, nb_iterations, seed):
    
    whisper_results = translate_better(audio)
    prompt = whisper_results[1]
    images = get_images(prompt)
    
    return whisper_results[0], whisper_results[1], images
    
#def diffuse(prompt, guidance_scale, nb_iterations, seed):
#    
#    generator = torch.Generator(device=device).manual_seed(int(seed))
#    
#    print("""
#    β€”
#    Sending prompt to Stable Diffusion ... 
#    β€”
#    """)
#    print("prompt: " + prompt)
#    print("guidance scale: " + str(guidance_scale))
#    print("inference steps: " + str(nb_iterations))
#    print("seed: " + str(seed))
#    
#    images_list = pipe(
#            [prompt] * 2, 
#            guidance_scale=guidance_scale,
#            num_inference_steps=nb_iterations, 
#            generator=generator
#        )
#    
#    images = []
#    
#    safe_image = Image.open(r"unsafe.png")
#    
#    for i, image in enumerate(images_list["sample"]):
#        if(images_list["nsfw_content_detected"][i]):
#            images.append(safe_image)
#        else:
#            images.append(image)
#
#    
#    print("Stable Diffusion has finished")
#    print("β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”")
#    
#    return images

def translate_better(audio):
    print("""
    β€”
    Sending audio to Whisper ...
    β€”
    """)
    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("β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”")  
    print("translated: " + translate_text_result)

    return transcribe_text_result, translate_text_result


#def translate(audio):
#    print("""
#    β€”
#    Sending audio to Whisper ...
#    β€”
#    """)
#    # current dateTime
#    now = datetime.now()    
#    # convert to string
#    date_time_str = now.strftime("%Y-%m-%d %H:%M:%S")
#    print('DateTime String:', date_time_str)
#    
#    audio = whisper.load_audio(audio)
#    audio = whisper.pad_or_trim(audio)
#    
#    mel = whisper.log_mel_spectrogram(audio).to(whisper_model.device)
#    
#    _, probs = whisper_model.detect_language(mel)
#    
#    transcript_options = whisper.DecodingOptions(task="transcribe", fp16 = False)
#    translate_options = whisper.DecodingOptions(task="translate", fp16 = False)
#    
#    transcription = whisper.decode(whisper_model, mel, transcript_options)
#    translation = whisper.decode(whisper_model, mel, translate_options)
#    
#    print("language spoken: " + transcription.language)
#    print("transcript: " + transcription.text)
#    print("β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”")  
#    print("translated: " + translation.text)
#    if transcription.language == "en":
#        tr_flag = flag.flag('GB')
#    else:
#        tr_flag = flag.flag(transcription.language)    
#    return tr_flag, transcription.text, translation.text

### β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”

css = """
        .container {
            max-width: 780px;
            margin: auto;
            padding-top: 1.5rem;
        }
        a {
            text-decoration: underline;
        }
        h1 {
            font-weight: 900; 
            margin-bottom: 7px;
            text-align: center;
            font-size: 2em;
            margin-bottom: 1em;
        }
        #w2sd_container{
            margin-top: 20px;
        }
        .footer {
            margin-bottom: 45px;
            margin-top: 35px;
            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;
        }
        .tabitem {
            border-bottom-left-radius: 10px;
            border-bottom-right-radius: 10px;
        }
        #record_tab, #upload_tab {
            font-size: 1.2em;
        }
        #record_btn{
            
        }
        #record_btn > div > button > span {
            width: 2.375rem;
            height: 2.375rem;
        }
        #record_btn > div > button > span > span {
            width: 2.375rem;
            height: 2.375rem;
        }
        audio {
            margin-bottom: 10px;
        }
        div#record_btn > .mt-6{
            margin-top: 0!important;
        }
        div#record_btn > .mt-6 button {
            font-size: 2em;
            width: 100%;
            padding: 20px;
            height: 160px;
        }
        div#upload_area {
            height: 11.1rem;
        }
        div#upload_area > div.w-full > div {
            min-height: 9rem;
        }
        #check_btn_1, #check_btn_2{
            color: #fff;
            --tw-gradient-from: #4caf50;
            --tw-gradient-stops: var(--tw-gradient-from), var(--tw-gradient-to);
            --tw-gradient-to: #4caf50;
            border-color: #8bc34a;
        }
        #magic_btn_1, #magic_btn_2{
            color: #fff;
            --tw-gradient-from: #f44336;
            --tw-gradient-stops: var(--tw-gradient-from), var(--tw-gradient-to);
            --tw-gradient-to: #ff9800;
            border-color: #ff9800;
        }
        input::-webkit-inner-spin-button, input::-webkit-outer-spin-button {
            -webkit-appearance: none;
        }
        input[type=number] {
            -moz-appearance: textfield;
        }
        input[type=range] {
            -webkit-appearance: none;
            cursor: pointer;
            height: 1px;
            background: currentColor;
        }
        input[type=range]::-webkit-slider-thumb {
            -webkit-appearance: none;
            width: 0.5em;
            height: 1.2em;
            border-radius: 10px;
            background: currentColor;
        }
        input[type=range]::-moz-range-thumb{
            width: 0.5em;
            height: 1.2em;
            border-radius: 10px;
            background: currentColor;
        }
        div#spoken_lang textarea {
            font-size: 4em;
            line-height: 1em;
            text-align: center;
        }
        div#transcripted {
            flex: 4;
        }
        div#translated textarea {
            font-size: 1.5em;
            line-height: 1.25em;
        }
        #sd_settings {
            margin-bottom: 20px;
        }
        #diffuse_btn {
            color: #fff;
            font-size: 1em;
            margin-bottom: 20px;
            --tw-gradient-from: #4caf50;
            --tw-gradient-stops: var(--tw-gradient-from), var(--tw-gradient-to);
            --tw-gradient-to: #4caf50;
            border-color: #8bc34a;
        }
        #notice {
            padding: 20px 14px 10px;
            display: flex;
            align-content: space-evenly;
            gap: 20px;
            line-height: 1em;
            font-size: .8em;
            border: 1px solid #374151;
            border-radius: 10px;
        }
        #about {
            padding: 20px;
        }
        #notice > div {
            flex: 1;
        }
        
"""   

### β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”β€”
 
with gr.Blocks(css=css) as demo:
    with gr.Column():
        gr.HTML('''
            <h1>
                Whisper to Stable Diffusion
            </h1>
            <p style='text-align: center;'>
                Ask stable diffusion for images by speaking (or singing πŸ€—) in your native language ! Try it in French πŸ˜‰
            </p>
            
            <p style='text-align: center;'>
                This demo is wired to the official SD Space β€’ Offered by Sylvain <a href='https://twitter.com/fffiloni' target='_blank'>@fffiloni</a> β€’ <img id='visitor-badge' alt='visitor badge' src='https://visitor-badge.glitch.me/badge?page_id=gradio-blocks.whisper-to-stable-diffusion' style='display: inline-block' /><br />
                β€”         
            </p>
    
        ''')
#        with gr.Row(elem_id="w2sd_container"):
#            with gr.Column():
            
        gr.Markdown(
            """
             
            ## 1. Record audio or Upload an audio file:
            """
        )
        
        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 ? πŸ‘", elem_id="check_btn_1")              
                    audio_r_direct_sd = gr.Button("Magic Whisper β€Ί SD right now!", elem_id="magic_btn_1")
        
        with gr.Tab(label="Upload audio input", elem_id="upload_tab"):
            with gr.Column():
                upload_input = gr.Audio(
                                    source="upload",
                                    type="filepath",
                                    show_label=False,
                                    elem_id="upload_area"
                                )
                with gr.Row():
                    audio_u_translate = gr.Button("Check Whisper first ? πŸ‘", elem_id="check_btn_2")              
                    audio_u_direct_sd = gr.Button("Magic Whisper β€Ί SD right now!", elem_id="magic_btn_2")
        
        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, correct it if necessary:
            """
        )
        
        with gr.Row():
            
            transcripted_output = gr.Textbox(
                                    label="Transcription in your detected spoken language", 
                                    lines=3,
                                    elem_id="transcripted"
                                )
            #language_detected_output = gr.Textbox(label="Native language", elem_id="spoken_lang",lines=3)
            
        with gr.Column():
            translated_output = gr.Textbox(
                                    label="Transcript translated in English by Whisper", 
                                    lines=4,
                                    elem_id="translated"
                                )
            with gr.Row():
                clear_btn = gr.Button(value="Clear")
                diffuse_btn = gr.Button(value="OK, Diffuse this prompt !", elem_id="diffuse_btn")
                
                clear_btn.click(fn=lambda value: gr.update(value=""), inputs=clear_btn, outputs=translated_output)
    
                
                
                
                    
#            with gr.Column():
                
                
                    
        gr.Markdown("""
            ## 3. Wait for Stable Diffusion Results β˜•οΈ
            Inference time is about ~20-30 seconds, when it's your turn 😬
            """
            ) 
        
        sd_output = gr.Gallery().style(grid=2, height="auto")
                
                
        gr.Markdown("""
            ### πŸ“Œ About the models
            <p style='font-size: 1em;line-height: 1.5em;'>   
            <strong>Whisper</strong> is a general-purpose speech recognition model.<br /><br />
            It is trained on a large dataset of diverse audio and is also a multi-task model that can perform multilingual speech recognition as well as speech translation and language identification. <br />
            β€”
            </p>
            <p style='font-size: 1em;line-height: 1.5em;'>
            <strong>Stable Diffusion</strong> is a state of the art text-to-image model that generates images from text.
            </p>
            <div id="notice">
                <div>
                LICENSE 
                <p style='font-size: 0.8em;'> 
                The model is licensed with a <a href="https://huggingface.co/spaces/CompVis/stable-diffusion-license" target="_blank">CreativeML Open RAIL-M</a> license.</p>
                <p style='font-size: 0.8em;'>  
                The authors claim no rights on the outputs you generate, you are free to use them and are accountable for their use which must not go against the provisions set in this license.</p>
                <p style='font-size: 0.8em;'>  
                The license forbids you from sharing any content that violates any laws, produce any harm to a person, disseminate any personal information that would be meant for harm, spread misinformation and target vulnerable groups.</p>
                <p style='font-size: 0.8em;'>  
                For the full list of restrictions please <a href="https://huggingface.co/spaces/CompVis/stable-diffusion-license" target="_blank" target="_blank">read the license</a>.
                 </p>
                 </div>
                 <div>
                 Biases and content acknowledgment
                 <p style='font-size: 0.8em;'>
                 Despite how impressive being able to turn text into image is, beware to the fact that this model may output content that reinforces or exacerbates societal biases, as well as realistic faces, pornography and violence.</p>
                <p style='font-size: 0.8em;'>  
                The model was trained on the <a href="https://laion.ai/blog/laion-5b/" target="_blank">LAION-5B dataset</a>, which scraped non-curated image-text-pairs from the internet (the exception being the removal of illegal content) and is meant for research purposes.</p>
                <p style='font-size: 0.8em;'>  You can read more in the <a href="https://huggingface.co/CompVis/stable-diffusion-v1-4" target="_blank">model card</a>.
                 </p>
                 </div>
             </div>
    
        """, elem_id="about")
                
        audio_r_translate.click(translate_better, 
                                inputs = record_input, 
                                outputs = [
                                    #language_detected_output,
                                    transcripted_output,
                                    translated_output
                                ])
        
        audio_u_translate.click(translate_better,
                                inputs = upload_input, 
                                outputs = [
                                    #language_detected_output,
                                    transcripted_output,
                                    translated_output
                                ]) 
        
        audio_r_direct_sd.click(magic_whisper_to_sd, 
                                inputs = [
                                    record_input, 
                                    guidance_scale, 
                                    nb_iterations, 
                                    seed
                                ], 
                                outputs = [
                                    #language_detected_output,
                                    transcripted_output,
                                    translated_output,
                                    sd_output
                                ])
        
        audio_u_direct_sd.click(magic_whisper_to_sd, 
                                inputs = [
                                    upload_input,
                                    guidance_scale,
                                    nb_iterations,
                                    seed
                                ], 
                                outputs = [
                                    #language_detected_output,
                                    transcripted_output,
                                    translated_output,
                                    sd_output
                                ])
        
        diffuse_btn.click(get_images, 
                              inputs = [
                                  translated_output
                                  ], 
                              outputs = sd_output
                          )
        gr.HTML('''
                <div class="footer">
                    <p>Whisper by <a href="https://github.com/openai/whisper" target="_blank">OpenAI</a> - Stable Diffusion by <a href="https://huggingface.co/CompVis" target="_blank">CompVis</a> and <a href="https://huggingface.co/stabilityai"  target="_blank">Stability AI</a>
                    </p>
                </div>
                ''')
        
    
if __name__ == "__main__":
    demo.queue(max_size=32, concurrency_count=20).launch()