ai enhance update, better scaler, more tests. better defaults

.gitignore

@@ -1,4 +1,5 @@
 .env
+test-*/
 
 # Byte-compiled / optimized / DLL files
 __pycache__/

config/default_sd15_conf.json

@@ -1,10 +1,13 @@
 {
   "sampler_name": "DPM++ 2M Karras",
+  "refiner_switch_at": 0.0,  
   "steps": 30,
   "cfg_scale": 7.0,
   "hr_upscaler": "Latent",
   "denoising_strength": 0.68,
   "subseed_strength": 0.75,
+  "restore_faces": false,
+  "negative_prompt": "bad_prompt_version2-neg",
   "override_settings_restore_afterwards": false,
   "override_settings": {
     "sd_model_checkpoint": "realisticVisionV60B1_v51VAE.safetensors",

config/default_sdxl_conf.json

@@ -7,6 +7,8 @@
   "hr_upscaler": "Latent",
   "denoising_strength": 0.68,
   "subseed_strength": 0.75,
+  "restore_faces": false,
+  "negative_prompt": "ac_neg1 ac_neg2",
   "override_settings_restore_afterwards": false,
   "override_settings": {
     "sd_model_checkpoint": "opendalle_v11.safetensors",

config/default_sdxl_lightning_conf.json

@@ -1,12 +1,13 @@
 {
   "sampler_name": "DPM++ SDE",
-  "refiner_checkpoint": null,
   "refiner_switch_at": 0.0,
   "steps": 4,
   "cfg_scale": 1.0,
   "hr_upscaler": "Latent",
   "denoising_strength": 0.68,
   "subseed_strength": 0.75,
+  "restore_faces": false,
+  "negative_prompt": "ac_neg1 ac_neg2",
   "override_settings_restore_afterwards": false,
   "override_settings": {
     "sd_model_checkpoint": "dreamshaperXL_lightningDPMSDE.safetensors",

images.py

@@ -45,7 +45,7 @@ async def OpenDallePrompt(prompt):
 7. **Additional details (5-10 words).** Include specific or important details that will complement and clarify the visualization.
 8. **Overall mood or impression (3-7 words).** Conclude with a description of the overall mood or main impression that the image should convey.
 
-Write everything in 1 sentence (not a list), separating with commas, in English, strictly following the instruction (especially the number of words)."""},
+Write everything in 1 sentence (not a list), separating with commas, in English, strictly following the instruction (especially the number of words). Provide descriptions that are direct and literal, avoiding metaphors and figurative language, and exclude specific features like "Velvet shadows." Focus on explicit and clear depiction of features, surroundings, element combinations, symbolic significance, lighting, additional details, and overall mood without using commonly associated expressions or imagery."""},
         {'role': 'assistant', 'content': "Understood, provide the input and I will generate a complete description of a scene."},
         {'role': 'user', 'content': "photo of an ancient castle very atmospheric, but the sky should be completely covered in clouds, yet it should be bright."},
         {'role': 'assistant', 'content': "photo of an ancient castle with a majestic, eerie ambiance, highlighting its towering spires, weathered stone, and ivy-clad walls, set against a backdrop of a dense, mystic forest under a sky completely shrouded in clouds yet illuminated by a diffused, ethereal light, where the fusion of natural decay and enduring architectural grandeur evokes a sense of timeless mystery and the eternal battle between man and nature, casting a glow that reveals subtle details and textures, creating an impression of haunting beauty and solemn tranquility. "},
@@ -79,7 +79,8 @@ def load_default_conf(self, filename):
     def maybe_scaler(self, payload, width, height):
         scale = math.sqrt(width * height) / self.base_model_size
 
-        if scale >= 1.2:
+        # If the target is more than 20% off the ideal size, scale it
+        if abs(scale - 1) >= 0.2:
             # SD expects pixel sized aligned by 8
             sd_width = 8 * round(width / (scale * 8))
             sd_height = 8 * round(height / (scale * 8))
@@ -158,7 +159,7 @@ async def generations(request: GenerationsRequest):
         if not request.prompt.startswith("I NEED to test how the tool works with extremely simple prompts. DO NOT add any detail, just use it AS-IS:"):
             request.prompt = revised_prompt = await OpenDallePrompt(request.prompt)
         else:
-            request.prompt = request.prompt[len("I NEED to test how the tool works with extremely simple prompts. DO NOT add any detail, just use it AS-IS:"):]
+            request.prompt = revised_prompt = request.prompt[len("I NEED to test how the tool works with extremely simple prompts. DO NOT add any detail, just use it AS-IS:"):]
 
     req = rg.create_request(request.prompt, int(width), int(height), request.n)
     print(req)

test/test_images-dall-e-1.md

@@ -1,20 +1,20 @@
 # A cute baby sea otter
-## dall-e-1 256x256 took 3.1 seconds
-![A cute baby sea otter](test_image_dall-e-1_256x256.png)
+## dall-e-1 256x256 took 36.4 seconds
+![A cute baby sea otter](test_image_dall-e-1_256x256-01_01.png)
 
 
 --------------------------------------------------
 
 
-## dall-e-1 512x512 took 3.4 seconds
-![A cute baby sea otter](test_image_dall-e-1_512x512.png)
+## dall-e-1 512x512 took 3.1 seconds
+![A cute baby sea otter](test_image_dall-e-1_512x512-01_01.png)
 
 
 --------------------------------------------------
 
 
-## dall-e-1 1024x1024 took 12.1 seconds
-![A cute baby sea otter](test_image_dall-e-1_1024x1024.png)
+## dall-e-1 1024x1024 took 12.2 seconds
+![A cute baby sea otter](test_image_dall-e-1_1024x1024-01_01.png)
 
 
 --------------------------------------------------

test/test_images-dall-e-2.md

@@ -1,20 +1,20 @@
 # A cute baby sea otter
-## dall-e-2 256x256 took 13.3 seconds
-![A cute baby sea otter](test_image_dall-e-2_256x256.png)
+## dall-e-2 256x256 took 32.2 seconds
+![A cute baby sea otter](test_image_dall-e-2_256x256-01_01.png)
 
 
 --------------------------------------------------
 
 
-## dall-e-2 512x512 took 2.7 seconds
-![A cute baby sea otter](test_image_dall-e-2_512x512.png)
+## dall-e-2 512x512 took 8.5 seconds
+![A cute baby sea otter](test_image_dall-e-2_512x512-01_01.png)
 
 
 --------------------------------------------------
 
 
-## dall-e-2 1024x1024 took 3.8 seconds
-![A cute baby sea otter](test_image_dall-e-2_1024x1024.png)
+## dall-e-2 1024x1024 took 5.4 seconds
+![A cute baby sea otter](test_image_dall-e-2_1024x1024-01_01.png)
 
 
 --------------------------------------------------

test/test_images-dall-e-3.md

@@ -1,23 +1,47 @@
 # A cute baby sea otter
-## dall-e-3 1024x1024 took 62.4 seconds
-![A cute baby sea otter](test_image_dall-e-3_1024x1024.png)
-revised_prompt: A cuddly, playful baby sea otter with soft, dark brown fur, large, curious eyes, and tiny, webbed paws floating on its back among lush kelp forests, basking in the gentle sunlight filtering through the crystal-clear water, exuding an irresistible charm and innocence that warms the heart, capturing the joy and wonder of nature's most adorable marine creatures.
+## dall-e-3 1024x1024 took 74.9 seconds
+![A cute baby sea otter](test_image_dall-e-3_1024x1024-01_01.png)
+revised_prompt:  A cute baby sea otter with round, fluffy fur, large, curious eyes, and tiny paws clutching a piece of kelp, floating on its back in a calm, serene ocean environment surrounded by lush, swaying kelp forests, embodying the innocence and playfulness of nature, under soft, warm sunlight filtering through the water's surface, showcasing the otter's endearing features and the gentle ripples of water around it, conveying a sense of joy and tranquility.
 
 
 --------------------------------------------------
 
 
-## dall-e-3 1024x1796 took 112.5 seconds
-![A cute baby sea otter](test_image_dall-e-3_1024x1796.png)
-revised_prompt: A cuddly, fluffy baby sea otter floating on its back, displaying its adorable button nose, sparkling eyes, and soft fur, surrounded by vibrant kelp forests and playful bubbles, with the sun casting a warm, golden glow on the crystal-clear water, embodying the innocence and joy of life, and showcasing the harmonious coexistence of marine creatures and their environment, capturing a delightful and heartwarming moment in nature.
+## dall-e-3 1024x1024 took 29.3 seconds
+![I NEED to test how the tool works with extremely simple prompts. DO NOT add any detail, just use it AS-IS:A cute baby sea otter](test_image_dall-e-3_1024x1024-not-enhanced-01_01.png)
+revised_prompt: A cute baby sea otter
 
 
 --------------------------------------------------
 
 
-## dall-e-3 1796x1024 took 112.0 seconds
-![A cute baby sea otter](test_image_dall-e-3_1796x1024.png)
-revised_prompt:  A cuddly, fluffy baby sea otter floating on its back, showcasing its soft, velvety fur, button-like eyes, and playful expression, surrounded by calm, crystal-clear waters and lush kelp forests, embodying the perfect blend of innocence and curiosity, as the sunlight dances on the water's surface, casting a warm, gentle glow that accentuates the otter's endearing features and captures the essence of pure joy and wonder, invoking feelings of delight and affection.
+## dall-e-3 1024x1796 took 111.4 seconds
+![A cute baby sea otter](test_image_dall-e-3_1024x1796-01_01.png)
+revised_prompt: A cute baby sea otter floating on its back, displaying its round, fluffy body, button-like eyes, and small, playful paws, clutching a piece of kelp for stability, surrounded by calm, crystal-clear waters and lush underwater vegetation, embodying the essence of innocence and curiosity, bathed in soft, warm sunlight filtering through the water's surface, with tiny air bubbles escaping from its nose, evoking feelings of delight and endearment.
+
+
+--------------------------------------------------
+
+
+## dall-e-3 1024x1796 took 85.1 seconds
+![I NEED to test how the tool works with extremely simple prompts. DO NOT add any detail, just use it AS-IS:A cute baby sea otter](test_image_dall-e-3_1024x1796-not-enhanced-01_01.png)
+revised_prompt: A cute baby sea otter
+
+
+--------------------------------------------------
+
+
+## dall-e-3 1796x1024 took 111.5 seconds
+![A cute baby sea otter](test_image_dall-e-3_1796x1024-01_01.png)
+revised_prompt: A cute baby sea otter with a round, fluffy body, playful dark eyes, and small, webbed paws, floating on its back in a calm, shallow ocean cove surrounded by lush kelp forests, its fur glistening with droplets of water as it curiously manipulates a seashell, embodying the innocence and wonder of nature, under soft, warm sunlight filtering through the surface, revealing the otter's delightful expression and intricate fur patterns, capturing an adorable and heartwarming moment.
+
+
+--------------------------------------------------
+
+
+## dall-e-3 1796x1024 took 84.4 seconds
+![I NEED to test how the tool works with extremely simple prompts. DO NOT add any detail, just use it AS-IS:A cute baby sea otter](test_image_dall-e-3_1796x1024-not-enhanced-01_01.png)
+revised_prompt: A cute baby sea otter
 
 
 --------------------------------------------------

test_images.py

@@ -0,0 +1,79 @@
+#!/usr/bin/env python
+import base64
+import time
+import argparse
+import sys
+import io
+from PIL import Image
+import openai
+
+client = openai.Client(base_url='http://localhost:5005/v1')
+
+TEST_DIR = 'test'
+not_enhanced = "I NEED to test how the tool works with extremely simple prompts. DO NOT add any detail, just use it AS-IS:"
+
+def generate_image(prompt, model, res, f, n = 1, suffix=''):
+    start = time.time()
+    response = client.images.generate(prompt=prompt, model=model, size=res, response_format='b64_json', n=n)
+    #image = Image.open(io.BytesIO(base64.b64decode(response.data[0].b64_json)))
+    #image.show()
+    end = time.time()
+    print(f"## {model} {res} took {end-start:.1f} seconds", file=f)
+
+    for i, img in enumerate(response.data, 1):
+        fname = f"test_image_{model}_{res}{suffix}-{i:02d}_{n:02d}.png"
+        with open(f'{TEST_DIR}/{fname}', 'wb') as png:
+            png.write(base64.b64decode(img.b64_json))
+        # markdown record the details of the test, including any extra revised_prompt
+        print(f"![{prompt}]({fname})", file=f)
+        if img.revised_prompt:
+            print("revised_prompt: " + img.revised_prompt, file=f)
+        print("\n", file=f, flush=True)
+
+    print("-"*50, file=f)
+    print("\n", file=f, flush=True)
+
+def full_test(prompt, n=1):
+    for model in ['dall-e-1', 'dall-e-2']:
+        with open(f"{TEST_DIR}/test_images-{model}.md", "w") as f:
+            print(f"# {prompt}", file=f)
+            for res in ['256x256', '512x512', '1024x1024']:
+                generate_image(prompt, model, res, f, n=n)
+
+    model = 'dall-e-3'
+    with open(f"{TEST_DIR}/test_images-{model}.md", "w") as f:
+        print(f"# {prompt}", file=f)
+        for res in ['1024x1024', '1024x1796', '1796x1024']:
+            generate_image(prompt, model, res, f, n=n)
+            generate_image(not_enhanced + prompt, model, res, f, n=n, suffix='-not-enhanced')
+
+def quick_test(prompt, n=1):
+    with open(f"{TEST_DIR}/test_images_quick.md", "w") as f:
+        print(f"# {prompt}", file=f)
+        generate_image(prompt, "dall-e-1", "512x512", f, n=n)
+        generate_image(prompt, "dall-e-1", "1024x1024", f, n=n)
+        generate_image(prompt, "dall-e-2", "1024x1024", f, n=n)
+        generate_image(not_enhanced + prompt, "dall-e-3", "1024x1024", f, n=n, suffix='-not-enhanced')
+        generate_image(prompt, "dall-e-3", "1024x1024", f, n=n)
+
+def parse_args(argv=None):
+    parser = argparse.ArgumentParser(formatter_class=argparse.ArgumentDefaultsHelpFormatter)
+
+    parser.add_argument('-p', '--prompt', action='store', type=str, default="A cute baby sea otter")
+    parser.add_argument('-q', '--quick', action='store_true')
+    parser.add_argument('-f', '--full', action='store_true')
+    parser.add_argument('-n', '--batch', action='store', type=int, default=1)
+
+    return parser.parse_args()
+
+
+if __name__ == '__main__':
+    args = parse_args(sys.argv[1:])
+
+    if args.quick:
+        quick_test(args.prompt, n=args.batch)
+    elif args.full:
+        full_test(args.prompt, n=args.batch)
+
+
+