🎨 AI Image Generator – Create Images in Scratch #
The AI Image Generator extension brings powerful AI image creation directly into Scratch.
Create custom sprites and backdrops using simple text descriptions with:
-
- OpenAI
- Google Gemini
- Stability AI
No drawing skills required ✨
🌟 Overview #
- Text to Image: Describe what you want and AI creates it instantly.
- Multiple AI Engines: OpenAI, Google Gemini, Stability AI.
- 13 Art Styles: Cartoon, pixel, 3D, watercolor, anime, fantasy, and more.
- Sprites and Backdrops: Generate costumes or full stage backgrounds.
- Type Guidance: Pick costume type (character/object) or backdrop type (outdoor/indoor/abstract/space/underwater) to steer results.
- Background Control: Automatically keep or remove costume backgrounds.
- Child-Safe: Built-in content moderation for kids.
- Flexible Pricing: Use free tiers, free credits, or paid usage depending on the engine.
✨ Key Features #
- Supports OpenAI DALL·E, Google Gemini/Imagen, and Stability AI image models.
- Tailored default models are included; use update model list to discover additional or newer ones.
- 13 predefined art styles optimized for Scratch projects.
- Costume and backdrop type presets to guide prompts (character/object, outdoor/indoor/abstract/space/underwater).
- Smart background removal for transparent sprite costumes.
- Automatic image resizing for sprites and stage backdrops.
- Sequential naming (ai-costume1, ai-backdrop1, …).
- Built-in safety moderation for kid-friendly prompts.
- Prompt quality boosters for better results.
- Separate API keys per engine.
🚀 How to Use #
- Go to: pishi.ai/play
- Open the Extensions panel.
- Select: مولّد الصور بالذكاء الاصطناعي
- Get an API key from one of the supported platforms:
Option 1: OpenAI (DALL·E)
- OpenAI API Keys
- Create a new API key (sk-proj-…).
- Billing information may be required.
Option 2: Google Gemini
- Google AI Studio
- Create an API key with your Google account (AIza…).
- Free tier available in many regions.
Tip: Free tier for image generation API access requires setting up Cloud Billing (which grants free usage credits).
Option 3: Stability AI (Recommended for Free Credits)
- Stability AI API Keys
- Create a free account.
- 🎁 New accounts receive free image generation credits.
- Generate an API key (sk-…) and start immediately.
Paste your key and start generating images.
🧱 Blocks and Functions #
🪄 Generate Sprite Costume #
Generates an AI image and adds it as a new sprite costume.
The image is resized for Scratch sprites and can optionally have a transparent background.
- PROMPT: Text description of what you want to generate (e.g. “cute robot”, “pixel spaceship”).
- MODEL: The AI image engine and model to use. For example:
[openai] dall-e-2, [gemini] gemini-2.5-flash-image, [stability] stable-image-core.
- To remove or keep backgrounds for future costume generations use: after costume generation [BG_MODE] the background
- The costume is added to the sprite that was selected when the block started.
- If no valid sprite exists, a new sprite is created automatically.
- The new costume is selected immediately.
Example:
🧩 Costume Type #
Sets the type of sprite costumes.
- COSTUME_TYPE: character or object
- Use character for characters (usually full-body) centered in frame.
- Use object for items or props with clean, isolated framing.
Tip: This setting stays active until you change it.
Example:
✂️ Background Removal (Costumes) #
Controls background removal for future costume generations.
- BG_MODE: remove (to make backgrounds transparent), keep (to keep backgrounds).
Tip: This setting only affects costumes, not backdrops.
🌄 Generate Backdrop #
Generates an AI image and adds it as a new stage backdrop.
The image is resized to fit the Scratch stage.
- PROMPT: Description of the scene or environment.
- MODEL: The AI image engine and model to use.
- Backdrops are resized to 480×360.
- The background is always kept.
- To guide the scene style use: set backdrop type to [BACKDROP_TYPE]
Example:
🏷️ Backdrop Type #
Sets the scene category used when generating backdrops.
- BACKDROP_TYPE: outdoor, indoor, abstract, space, or underwater.
Tip: This setting stays active until you change it.
Example:
🎭 Style Customization #
Sets the visual art style used for all future image generations.
- STYLE: One of the predefined art styles.
- cartoon, simple, pixel, 3d, watercolor, sketch
- comic, anime, origami, clay, voxel, sticker, fantasy
Tip: The selected style remains active until you change it.
Example:
🔑 API Key Management #
Stores an API key for the selected AI engine.
- ENGINE: Choose one of: openai, gemini, stability.
- Each engine’s key is stored separately.
- Keys are saved only in the current browser session.
- Keys are cleared when the tab is closed.
Tip: Stability AI is a good starting choice because it offers free credits on signup.
🤖 Model Management #
Fetches the latest available image models from the selected platform.
- ENGINE: The AI platform to update models for.
- Requires a valid API key.
- Calling this can add new models to the menu, including newly published models and aliases or earlier versions of default models.
- Example: nano-banana-pro-preview which is actually a preview/alias for gemini-3-pro-image-preview.
Default model list (before updates):
- OpenAI: dall-e-3, dall-e-2
- Google Gemini: gemini-2.5-flash-image, gemini-3-pro-image-preview, imagen-4.0-generate-001, imagen-4.0-fast-generate-001
- Stability AI: stable-diffusion-xl-1024-v1-0, stable-image-core
📊 Status and Errors #
Returns the name of the most recently generated asset.
- ASSET_TYPE: costume or backdrop
Returns the last image generation error message, or empty if successful.
- Useful for debugging and showing friendly messages to users.
🎓 Educational Uses #
- Understand how AI generates images:
Students explore the process of turning text into visuals and learn how modern multimodal models work. - Practice comparing AI engines and art styles:
Try different models (OpenAI, Stability, etc.) and see how each one creates unique results, even from the same prompt. - Create game assets instantly:
Make characters, backgrounds, props, and icons without needing drawing skills, making Scratch-style projects more accessible. - Learn prompt engineering through experimentation:
Students see how wording affects the final image and discover techniques for clarity, style control, and creativity. - Enhance storytelling and animation projects:
Quickly generate scenes and characters for interactive stories, comics, and animations. - Practice safe and responsible AI usage:
Discuss why some prompts are restricted, how moderation protects users, and why ethical AI matters in creative tools. - Encourage personal creativity and ownership:
Students can generate custom designs that match their own ideas, making learning more engaging and expressive. - Support teachers in producing visual materials:
Educators can generate diagrams, teaching aids, and simple illustrations directly inside pishi.ai projects.
🔧 Tips and Troubleshooting #
🧠 Choosing the Right Engine #
Which engine is best depends on your priorities: image quality, speed, cost, or control. Below is a concise comparison to help pick the right one for a classroom or project.
- Best overall fidelity and understanding: OpenAI DALL·E 3
- Fastest iterations & strong image quality: Google Gemini (Flash / fast Imagen variants)
- Most customization & artistic control: Stability AI (SDXL / stable-image-core / SD3)
- Best free starting credits: Stability AI (free credits on signup)
Quick selection tips: For classroom use start with Stability to save costs and experiment. For final high-quality assets use DALL·E 3. For very fast feedback loops (sketches, thumbnails) try Gemini Flash or the faster Imagen variants.
Notes:
- Use simple styles for cleaner background removal.
- Be specific in prompts for better results.
- Generate multiple costumes to animate sprites.
- Monitor usage to avoid unexpected costs.
🔒 Privacy and Safety #
- Internet connection required for image generation.
- Prompts are checked for kid-safe content.
- Images are stored locally in your Scratch project.
- Never share API keys publicly.
- Always supervise children when using AI tools.
💰 Costs and Usage #
OpenAI: Paid per image
Google Gemini: Free tier (requires setting up Cloud Billing) + paid options
Stability AI: Free credits on signup + credit-based pricing
- Use free tiers and credits for education.
- Start with cheaper models for testing.
- Write clear prompts to avoid regenerations.
🔗 Related Extensions #
- 💬 ChatGPT – generate better prompts
- 🎤 Speech Recognition – speak prompts
- 🖊️ Text to Speech – narrate images
- 📝 Translate – multilingual prompts
