Update app.py

app.py

@@ -24,31 +24,28 @@ from lxml.html import fromstring
 from diffusers.utils import export_to_gif, load_image
 from huggingface_hub import hf_hub_download
 from safetensors.torch import load_file, save_file
-from diffusers import FluxPipeline, DiffusionPipeline, AnimateDiffPipeline, MotionAdapter, EulerAncestralDiscreteScheduler, DDIMScheduler, StableDiffusionXLPipeline, UNet2DConditionModel, AutoencoderKL, UNet3DConditionModel
+from diffusers import AnimateDiffPipeline, MotionAdapter, DDIMScheduler, StableDiffusionPipeline
 
 # logging
 
 warnings.filterwarnings("ignore")
 root = logging.getLogger()
 root.setLevel(logging.DEBUG)
-handler = logging.StreamHandler(sys.stdout)
+handler = logging.StreamHandler(sys.stderr)
 handler.setLevel(logging.DEBUG)
 formatter = logging.Formatter('\n >>> [%(levelname)s] %(asctime)s %(name)s: %(message)s\n')
 handler.setFormatter(formatter)
 root.addHandler(handler)
-handler2 = logging.StreamHandler(sys.stderr)
-handler2.setLevel(logging.DEBUG)
-formatter = logging.Formatter('\n >>> [%(levelname)s] %(asctime)s %(name)s: %(message)s\n')
-handler2.setFormatter(formatter)
-root.addHandler(handler2)
 
 # constant data
 
-dtype = torch.float16
 if torch.cuda.is_available():
     device = "cuda"
+    dtype = torch.float16
 else:
     device = "cpu"
+    dtype = torch.float16
+
 base = "SG161222/Realistic_Vision_V6.0_B1_noVAE"
 adapter = MotionAdapter.from_pretrained("guoyww/animatediff-motion-adapter-v1-5-3", torch_dtype=dtype, device=device)
 
@@ -61,11 +58,10 @@ result = []
 
 seq=512
 fast=True
-fps=20
-time=3
+fps=18
 width=896
 height=896
-step=50
+step=30
 accu=8.5
 
 # ui data
@@ -112,7 +108,8 @@ function custom(){
 
 # torch pipes
 
-pipe = AnimateDiffPipeline.from_pretrained(base, motion_adapter=adapter).to(device)
+image_pipe = StableDiffusionPipeline.from_pretrained(base, torch_dtype=dtype, safety_checker=None).to(device)
+pipe = AnimateDiffPipeline.from_pretrained(base, torch_dtype=dtype, motion_adapter=adapter).to(device)
 pipe.scheduler = DDIMScheduler(
     clip_sample=False,
     beta_start=0.00085,
@@ -138,8 +135,8 @@ def xpath_finder(str,pattern):
 def translate(text,lang):
     if text == None or lang == None:
         return ""       
-    text = re.sub(f'[{punctuation}]', '', re.sub('[\s+]', ' ', text)).lower().strip()
-    lang = re.sub(f'[{punctuation}]', '', re.sub('[\s+]', ' ', lang)).lower().strip()    
+    text = re.sub(f'[{punctuation}]', '', re.sub('[ ]+', ' ', text)).lower().strip()
+    lang = re.sub(f'[{punctuation}]', '', re.sub('[ ]+', ' ', lang)).lower().strip()    
     if text == "" or lang == "":
         return ""
     if len(text) > 38:
@@ -151,7 +148,7 @@ def translate(text,lang):
         'Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/605.1.15 (KHTML, like Gecko) Version/16.1 Safari/605.1.15',
         'Mozilla/5.0 (Macintosh; Intel Mac OS X 13_1) AppleWebKit/605.1.15 (KHTML, like Gecko) Version/16.1 Safari/605.1.15'
     ]
-    padded_chars = re.sub("[(^\-)(\-$)]","",text.replace("","-").replace("- -"," ")).strip()
+    padded_chars = re.sub("(^-)|(-$)","",text.replace("","-").replace("- -"," ")).strip()
     query_text = f'Please translate {padded_chars}, into {lang}'
     url = f'https://www.google.com/search?q={query_text}'
     content = str(requests.get(
@@ -167,40 +164,43 @@ def translate(text,lang):
     trgt_text = xpath_finder(content,'//*[@id="tw-target-text"]/*')
     if trgt_lang == lang:
         translated = trgt_text
-    ret = re.sub(f'[{punctuation}]', '', re.sub('[\s+]', ' ', translated)).lower().strip()
-    print(ret)
+    ret = re.sub(f'[{punctuation}]', '', re.sub('[ ]+', ' ', translated)).lower().strip()
     return ret
     
 def generate_random_string(length):
     characters = str(ascii_letters + digits)
     return ''.join(random.choice(characters) for _ in range(length))
 
-def pipe_generate(img,p1,p2,motion):
+def pipe_generate(img,p1,p2,motion,time,title):
     global last_motion
     global pipe
 
-    if last_motion != motion:
-        if last_motion != "":
-            pipe.unload_lora_weights()
-        if motion != "":
-            pipe.load_lora_weights(motion, adapter_name="motion")
-            pipe.fuse_lora()
-            pipe.set_adapters("motion", [0.7])
-        last_motion = motion
-
-    pipe.to(device,dtype=dtype)
-
-    if img == None:
-        img = pipe(
+    if img is None:
+        img = image_pipe(
             prompt=p1,
+            negative_prompt=p2,
             height=height,
             width=width,
             guidance_scale=accu,
+            num_images_per_prompt=1,
             num_inference_steps=step,
             max_sequence_length=seq,
+            need_safetycheck=False,
             generator=torch.Generator(device).manual_seed(0)
         ).images[0]
 
+    if time == 0.0:
+        return img
+
+    if last_motion != motion:
+        if last_motion != "":
+            pipe.unload_lora_weights()
+        if motion != "":
+            pipe.load_lora_weights(motion, adapter_name="motion")
+            pipe.fuse_lora()
+            pipe.set_adapters("motion", [0.7])
+        last_motion = motion
+
     return pipe(
         prompt=p1,
         negative_prompt=p2,
@@ -210,32 +210,54 @@ def pipe_generate(img,p1,p2,motion):
         num_inference_steps=step,
         guidance_scale=accu,
         num_frames=(fps*time)
-    )
+    ).frames[0]
 
-def handle_generate(*inp):
+def handle_generate(*_inp):
 
-    inp = list(inp)
+    inp = list(_inp)
     
     inp[1] = translate(inp[1],"english")
     inp[2] = translate(inp[2],"english")
 
     if inp[2] != "":
-        inp[2] = f", {inp[2]}"
+        arr = []
+        for wrd in inp[2].split():
+            arr.append(wrd)
+        inp[2] = ", " + ", ".join(arr)
 
-    inp[2] = f"(deformed iris, deformed pupils, semi-realistic, cgi, 3d, render, sketch, cartoon, drawing, anime), text, cropped, out of frame, worst quality, low quality, jpeg artifacts, ugly, duplicate, morbid, mutilated, extra fingers, mutated hands, poorly drawn hands, poorly drawn face, mutation, deformed, blurry, dehydrated, bad anatomy, bad proportions, extra limbs, cloned face, disfigured, gross proportions, malformed limbs, missing arms, missing legs, extra arms, extra legs, fused fingers, too many fingers, long neck{inp[2]}"
+    inp[2] = f"creative, fake, error, dreamy, unreal, pixelated, bright, deformed iris, deformed pupils, semi-realistic, cgi, 3d, render, sketch, cartoon, drawing, anime, cropped, out of frame, worst quality, low quality, jpeg artifacts, ugly, duplicate, weird, morbid, mutilated, extra fingers, mutated hands, poorly drawn hands, poorly drawn face, mutation, deformed, blurry, dehydrated, bad anatomy, bad proportions, extra limbs, cloned face, disfigured, unspecified, general, gross proportions, malformed limbs, missing arms, missing legs, extra arms, extra legs, fused fingers, too many fingers, long neck{inp[2]}"
     
-    _do = ['photographed', 'realistic', 'dynamic poze', 'deep field', 'reasonable', "natural", 'rough', 'best quality', 'focused', "highly detailed"]
+    _do = ['photographed', "environment", 'realistic', 'true', 'genuine', 'dynamic poze', 'authentic', 'deep field', 'reasonable', "natural", 'rough', "real world", 'best quality', 'focused', "highly detailed"]
+
     if inp[1] != "":
-        _do.append(f"a new {inp[1]} content in the image")
+        for wrd in inp[1].split():
+            _do.append(wrd)
     inp[1] = ", ".join(_do)
+
+    if inp[5] != "":
+        _do.append(f'centered readable large bold written text: {inp[5]}')
+
+    inp[1] = ", ".join(_do)
+
+    print(f"""
+
+        - - - -
+        {inp[1]}
+        {inp[2]}
+        - - - -
+
+    """)
     
     ln = len(result)
     pipe_out = [pipe_generate(*inp) for i in range(ln)]
     
     names = []
     for i in pipe_out:
-        name = generate_random_string(12)+".png"
-        export_to_gif(i.frames[0],name,fps=fps)
+        name = generate_random_string(12) + ( ".png" if time == 0 else ".gif" )
+        if inp[4] == 0.0:
+            i.save(name)
+        else:
+            export_to_gif(i,name,fps=fps)
         names.append( name )
     return names
 
@@ -243,31 +265,45 @@ def ui():
     global result
     with gr.Blocks(theme=gr.themes.Soft(),css=css,js=js) as demo:
         gr.Markdown(f"""
-            # MULTI-LANGUAGE GIF CREATOR
+            # MULTI-LANGUAGE GIF/PNG CREATOR
         """)
         with gr.Row(elem_id="col-container"):
             with gr.Column():
                 with gr.Row():
-                    img = gr.Image(label="UPLOAD PHOTO",show_label=True,container=True,type="pil")
+                    img = gr.Image(label="Upload photo",show_label=True,container=False,type="pil")
+            with gr.Column(scale=0.66):
                 with gr.Row():
+                    title = gr.Textbox(
+                        placeholder="Logo title",
+                        container=False,
+                        max_lines=1
+                    )
                     prompt = gr.Textbox(
                         elem_id="prompt",
-                        placeholder="INCLUDE",
+                        placeholder="Included keywords",
                         container=False,
                         max_lines=1
                     )
                 with gr.Row():
                     prompt2 = gr.Textbox(
                         elem_id="prompt2",
-                        placeholder="EXCLUDE",
+                        placeholder="Excluded keywords",
                         container=False,
                         max_lines=1
                     )
-                with gr.Row(visible=False):
+                with gr.Row():
+                    time = gr.Slider(
+                        minimum=0.0,
+                        maximum=8.0,
+                        value=0.0,
+                        step=1.0,
+                        label="GIF/PNG Duration (0s = PNG)"
+                    )
+                with gr.Row():
                         motion = gr.Dropdown(
-                            label='CAMERA',
+                            label='GIF camera movement',
                             show_label=True,
-                            container=True,
+                            container=False,
                             choices=[
                                 ("(No Effect)", ""),
                                 ("Zoom in", "guoyww/#animatediff-motion-lora-zoom-in"),
@@ -282,11 +318,10 @@ def ui():
                             value="",
                             interactive=True
                         )
-            with gr.Column():
-                with gr.Row():
-                    run_button = gr.Button("START",elem_classes="btn",scale=0)
                 with gr.Row():
-                    result.append(gr.Image(interactive=False,elem_classes="image-container", label="Result", show_label=False, type='filepath', show_share_button=False))
+                    result.append(gr.Image(interactive=False,elem_classes="image-container", label="Result", show_label=True, type='filepath', show_share_button=False))
+        with gr.Row():
+            run_button = gr.Button("Start!",elem_classes="btn",scale=0)
 
         gr.on(
             triggers=[
@@ -295,7 +330,7 @@ def ui():
                 prompt2.submit
             ],
             fn=handle_generate,
-            inputs=[img,prompt,prompt2,motion],
+            inputs=[img,prompt,prompt2,motion,time,title],
             outputs=result
         )
         demo.queue().launch()
@@ -306,4 +341,4 @@ if __name__ == "__main__":
     os.chdir(os.path.abspath(os.path.dirname(__file__)))
     ui()
 
-# end
+# end