better magic prompts

app.py

@@ -17,12 +17,52 @@ if APP_PASSWORD is None:
 
 GENERATION_TIMEOUT = float(os.getenv("GENERATION_TIMEOUT", 60))
 
-def enhance_prompt(original_prompt):
+MAGIC_PROMPTS = [
+    """You are an expert DALL-E 2 image generation prompt optimizer. Your task is to take the user's initial prompt idea and enhance it to produce the best possible results from DALL-E 2. Follow these steps:
+Analyze the user's base prompt.
+Expand the prompt by adding specific details about:
+Subject/main elements
+Setting/background
+Lighting and atmosphere
+Color palette
+Artistic style (e.g. photorealistic, oil painting, digital art)
+Composition and perspective
+Incorporate descriptive adjectives and evocative language.
+Add relevant artistic/technical terms (e.g. macro shot, fisheye lens, chiaroscuro).
+Include quality boosters like "highly detailed", "award-winning", "stunning".
+Specify image type if relevant (e.g. digital illustration, 35mm photograph).
+Mention any desired emotions or moods.
+Avoid negative language - focus on what should be included rather than excluded.
+Keep text simple and minimal if text is required in the image.
+Aim for a prompt length of 40-60 words.
+Format the final prompt clearly, using proper punctuation and capitalization.
+Only return the improved prompt without any additional comments or messages.""",
+    """You are an expert DALL-E 3 image generation prompt optimizer. Enhance and refine anything the user sends you as an image generation prompt for DALL-E 3:
+
+Provide an improved version of the user message by following these guidelines:
+Adds more specific details about the scene, subjects, and atmosphere
+Incorporates precise descriptors for colors, textures, and lighting
+Specifies the artistic style or medium (e.g., oil painting, digital art, photography)
+Includes relevant compositional elements (foreground, background, perspective)
+Adds any missing context or setting information
+Removes any redundant or vague language
+Ensures the prompt is coherent and follows a logical structure
+Incorporates relevant technical terms or jargon related to art or photography
+Suggests any additional elements that could enhance the overall image
+Optimizes the prompt length for DALL-E 3's capabilities (aim for 40-60 words)
+Only return the improved prompt without any additional comments or messages.
+""",
+"Verbessere den folgenden Bildgenerierungsprompt, um ihn detaillierter und kreativer zu machen."
+]
+
+def enhance_prompt(original_prompt, img_model="dall-e-2"):
+    system_message = MAGIC_PROMPTS[0] if img_model == "dall-e-2" else MAGIC_PROMPTS[1]
+
     try:
         response = client.chat.completions.create(
             model="gpt-4o-mini",
             messages=[
-                {"role": "system", "content": "Verbessere den folgenden Bildgenerierungsprompt, um ihn detaillierter und kreativer zu machen."},
+                {"role": "system", "content": system_message},
                 {"role": "user", "content": original_prompt}
             ],
             max_tokens=300
@@ -33,26 +73,29 @@ def enhance_prompt(original_prompt):
         print("Ein Fehler ist beim Verbessern des Prompts aufgetreten:", e)
         return original_prompt
 
-def generate_image(prompt, use_magic_prompt, style, password, user_id, last_generation_time):
-    print(f"\nGenerierungsversuch von User {user_id} mit Prompt:\n{prompt}")
+def generate_image(prompt, use_magic_prompt, style, password, user_id, last_generation_time, model):
+    print(f"\nGenerierungsversuch von User {user_id} mit Modell {model} mit Prompt:\n{prompt}")
+
+    if prompt=="": 
+        return "Bitte gib einen Bildprompt ein.", None, last_generation_time
     
     if password != APP_PASSWORD:
         print("Generierung abgebrochen: Falsches Passwort.")
         return "Falsches Passwort. Bitte versuche es erneut.", None,  last_generation_time
 
     current_time = time.time()
-    if current_time - last_generation_time < GENERATION_TIMEOUT:
+    if model == "dall-e-3" and current_time - last_generation_time < GENERATION_TIMEOUT:
         remaining_time = max(0, int(GENERATION_TIMEOUT - (current_time - last_generation_time)))
         print("Generierung abgebrochen: Cooldown noch nicht abgelaufen")
         return f"Bitte warte noch {remaining_time} Sekunden bis zur nächsten Bildgenerierungen.\n\nNutze die Wartezeit, um deinen Prompt zu verfeinern. &#128521;", None,  last_generation_time
 
     if use_magic_prompt:
-        prompt = enhance_prompt(prompt)
+        prompt = enhance_prompt(prompt, model)
         print("\nFinaler Prompt:", prompt)
 
     try:
         response = client.images.generate(
-            model="dall-e-3",
+            model=model,
             prompt=prompt,
             size="1024x1024",
             quality="standard",
@@ -63,7 +106,7 @@ def generate_image(prompt, use_magic_prompt, style, password, user_id, last_gene
         image_url = response.data[0].url
         
         # Update the last generation time for this user
-        last_generation_time = current_time
+        last_generation_time = current_time if model == "dall-e-3" else last_generation_time
         
         return prompt, image_url, last_generation_time
     except Exception as e:
@@ -106,9 +149,9 @@ with gr.Blocks() as demo:
     user_id = gr.State(generate_user_id)
     last_generation_time = gr.State(0)
 
-    gr.Markdown("# codora DALL-E 3 Bildgenerator")
+    gr.Markdown("# codora DALL-E Bildgenerator")
     welcome_message = gr.Markdown("Willkommen!")
-    gr.Markdown(f"Gib einen Bildprompt ein und verwende optional die magische Prompt-Funktion, um ihn zu verbessern. Du kannst nur ein Bild alle {GENERATION_TIMEOUT} Sekunden generieren.")
+    gr.Markdown(f"Gib einen Bildprompt ein und verwende optional die magische Prompt-Funktion, um ihn zu verbessern. Mit DALL·E 3 kannst du ein Bild alle {GENERATION_TIMEOUT} Sekunden generieren. Mit DALL·E 2 kannst du so viele Bilder generieren, wie du möchtest.")
 
     with gr.Row():
         with gr.Column(scale=1):
@@ -119,6 +162,11 @@ with gr.Blocks() as demo:
                 choices=["vivid", "natural"],
                 value="vivid"
             )
+            model = gr.Radio(
+                label="Modell",
+                choices=["dall-e-2", "dall-e-3"],
+                value="dall-e-2"
+            )
             password = gr.Textbox(label="App Passwort", type="password")
             generate_button = gr.Button("Bild generieren")
 
@@ -134,7 +182,7 @@ with gr.Blocks() as demo:
 
     generate_button.click(
         fn=generate_image,
-        inputs=[prompt, use_magic_prompt, style, password, user_id, last_generation_time],
+        inputs=[prompt, use_magic_prompt, style, password, user_id, last_generation_time, model],
         outputs=[final_prompt, generated_image, last_generation_time]
     )