Generate and edit images. Saves the image and prints the exact path it wrote. Use when the user wants any visual creative output, an image generated from a description, or wants to modify an existing image in any way — editing, transforming, or adding visual elements.
Standardmäßig ist der Prompt ausgewählt, der zuerst die Quelle prüft. Sie können zu einem direkten Befehl wechseln oder eine lokale Kopie herunterladen.
Quelldateien prüfen
Lesen Sie SKILL.md und alle von SkillsMP angezeigten Begleitdateien, bevor Sie sich für eine Installation entscheiden.
Mit Codex oder Claude installieren Kopieren Sie diesen Prompt, fügen Sie ihn in Codex, Claude oder einen anderen Assistant ein und lassen Sie die Skill-Seite prüfen und installieren.
Ein direkter Befehl überspringt den Prüf-Prompt. Prüfen Sie die Quelle, bevor Sie ihn ausführen.
Generate and edit images. Saves the image and prints the exact path it wrote. Use when the user wants any visual creative output, an image generated from a description, or wants to modify an existing image in any way — editing, transforming, or adding visual elements.
Imagine – Image Generation & Editing via zdx imagine
Generate images from text prompts or edit existing images using Gemini, OpenAI, or Alibaba (Qwen-Image) image models. Supports text-to-image generation, image editing (inpainting, outpainting, style transfer), and multi-image composition.
CLI reference
zdx imagine --prompt <PROMPT> [OPTIONS]
Options:
-p, --prompt <PROMPT> Text prompt (required)
-s, --source <IMAGE> Source image for editing (repeatable for multi-image)
-o, --out <PATH> Output path — written exactly as given (default: $ZDX_HOME/artifacts/image-<timestamp>.<ext>)
--model <MODEL> Model override (default: gemini:gemini-3.1-flash-image-preview)
--aspect <RATIO> Aspect ratio (Gemini only; see table below)
--size <SIZE> 512px | 1K (default) | 2K | 4K
Output: prints the exact saved file path(s) to stdout. --out is honored literally; the model picks the image format (often JPEG for Gemini), so a file named .png may actually hold JPEG bytes. That is fine — zdx reads images by content, not extension. Use --out to control the path and always view/attach the printed path.
When running inside zdx (TUI, bot, or CLI), $ZDX_ARTIFACT_DIR is the preferred output location. Pass it via --out:
If no --out is given, images default to $ZDX_HOME/artifacts/.
Provider guidance
Default: If you do not pass --model, zdx imagine uses gemini:gemini-3.1-flash-image-preview.
OpenAI support:zdx imagine also supports both openai:gpt-image-2 and openai-codex:gpt-image-2.
Preferred OpenAI path: When using OpenAI, prefer openai-codex:gpt-image-2 unless the user explicitly asks for plain openai:gpt-image-2.
Alibaba (Qwen-Image) support:zdx imagine supports alibaba:qwen-image-3.0-pro, alibaba:qwen-image-3.0, and alibaba:qwen-image-2.0-pro — combined text-to-image + editing models on Alibaba DashScope. Requires ALIBABA_API_KEY. Use them when the user asks for Qwen/Alibaba image generation.
Preferred Alibaba path: Use alibaba:qwen-image-3.0-pro. It is the strongest of the three at dense layouts, small text (~10px), and multilingual typography. Fall back to alibaba:qwen-image-3.0 for cheaper/faster runs, and to alibaba:qwen-image-2.0-pro only when explicitly asked.
Gemini vs OpenAI/Alibaba:--aspect currently works with Gemini only; OpenAI, OpenAI Codex, and Alibaba require --size instead.
OpenAI size support: OpenAI/OpenAI Codex support 1K, 2K, and 4K. 512px is Gemini-only.
Alibaba size support: All Alibaba Qwen-Image models support 512px, 1K, and 2K (not 4K).
Recommended model selection
Use Gemini by default unless the user asks for OpenAI or there is a specific reason to use it.
Use openai-codex:gpt-image-2 as the preferred OpenAI model path.
Use openai:gpt-image-2 only when the user explicitly asks for the non-Codex OpenAI provider.
OpenAI examples
Preferred OpenAI Codex usage:
zdx imagine -p "A cinematic photo of a red fox in falling snow" --model openai-codex:gpt-image-2 --size 1K
Plain OpenAI usage:
zdx imagine -p "A cinematic photo of a red fox in falling snow" --model openai:gpt-image-2 --size 1K
OpenAI editing:
zdx imagine -p "Add a neon sign above the doorway, keep the rest unchanged" -s street.png --model openai-codex:gpt-image-2 --size 2K
Alibaba (Qwen-Image) examples
Text-to-image:
zdx imagine -p "A cinematic photo of a red fox in falling snow" --model alibaba:qwen-image-3.0-pro --size 1K
Editing (same model, add source image):
zdx imagine -p "Make the jacket bright red, keep everything else the same" -s person.png --model alibaba:qwen-image-3.0-pro --size 1K
Modes
Mode
Usage
Description
Text → Image
-p "prompt"
Generate from scratch
Edit single image
-p "prompt" -s image.png
Inpaint, outpaint, style transfer, modify
Multi-image compose
-p "prompt" -s a.png -s b.png
Combine images into new scene
Output location
Images are saved to $ZDX_ARTIFACT_DIR when set, otherwise $ZDX_HOME/artifacts/. Use --out to set the exact path:
Describe the scene, don't just list keywords. A narrative, descriptive paragraph will almost always produce a better, more coherent image than a list of disconnected words.
Infographic & visual explainer prompts
Template: Concept explainer
Create a [style] infographic that explains [concept/topic]. Show [key elements]
with clear labels and [visual metaphor or analogy]. The layout should be
[layout description]. Use [color palette] and [typography style].
[Aspect ratio].
Template: Process / how-it-works
A [style] diagram showing how [process/system] works, step by step.
Label each stage clearly: [stage 1], [stage 2], [stage 3].
Use [visual style: arrows, flow, numbered panels]. [Color scheme].
[Aspect ratio].
Template: Comparison / versus
A [style] side-by-side comparison of [thing A] vs [thing B].
Each side shows [key differences] with clear labels.
Use [visual contrast: split layout, color coding]. [Aspect ratio].
Template: Anatomy / breakdown
A [style] annotated breakdown of [subject], with labeled parts showing
[components/layers]. [Drawing style: technical, Da Vinci sketch, blueprint,
cutaway, cross-section]. Notes in English. [Aspect ratio].
Template: Timeline / history
A [style] visual timeline of [topic], from [start] to [end].
Key milestones labeled with dates and brief descriptions.
[Layout: horizontal, vertical, spiral]. [Color palette]. [Aspect ratio].
Template: Mental model / analogy
A [style] visual that explains [abstract concept] using the analogy of
[concrete thing]. Show how [mapping between concept and analogy].
Clear labels connecting the analogy to the real concept. [Aspect ratio].
Tips for better explainer images
State the educational goal — "explain X to someone unfamiliar" produces more accessible results than just naming the topic.
Describe the layout — triptych, grid, flowchart, numbered panels, split-screen. The model follows layout instructions well.
Use analogies and metaphors — "explain photosynthesis as a recipe" or "explain TCP as a postal system" produces creative, memorable visuals.
Specify label and text placement — "with bold labels", "numbered steps", "annotated arrows pointing to each part".
Choose a visual style that fits the content: blueprint for technical, watercolor for organic, flat vector for clean/modern, sketch for conceptual.
Use 16:9 or 3:2 for most explainers (good for horizontal layouts with room for labels).
Example prompts
How something works:
zdx imagine -p "A colorful, educational infographic explaining how a CPU executes an instruction. Show the fetch-decode-execute cycle as a circular flow diagram with labeled stages. Each stage has a small illustration: memory fetching data, decoder breaking it down, ALU computing. Flat vector style, vibrant colors on dark background." --aspect 16:9
Concept analogy:
zdx imagine -p "A whimsical illustrated infographic explaining the human immune system as a medieval castle defense. White blood cells as knights, antibodies as archers on walls, the skin as castle walls, fever as pouring boiling oil. Labeled annotations connecting each metaphor to the real biology. Colorful storybook illustration style." --aspect 16:9
Anatomy / breakdown:
zdx imagine -p "Da Vinci style anatomical sketch of a dissected Monarch butterfly. Detailed drawings of the head, wings, and legs on textured parchment with handwritten notes in English explaining each part." --aspect 1:1
Comparison:
zdx imagine -p "A clean, modern split-screen comparison of REST vs GraphQL APIs. Left side shows REST with multiple endpoint arrows, right side shows GraphQL with a single endpoint and flexible query. Color-coded: blue for REST, purple for GraphQL. Flat design, bold labels, white background." --aspect 16:9
Process / step-by-step:
zdx imagine -p "A vibrant infographic explaining photosynthesis as a recipe from a colorful kids' cookbook. Show the 'ingredients' (sunlight, water, CO2) going into a plant 'kitchen' and the 'finished dish' (sugar/energy) coming out. Numbered steps, playful illustrations, bright colors. Suitable for a 4th grader." --aspect 16:9
Timeline:
zdx imagine -p "An illustrated horizontal timeline of the history of programming languages, from Fortran (1957) to Rust (2015). Each milestone shows the language name, year, and a small icon representing its key innovation. Retro-futuristic style with muted earth tones and clean typography." --aspect 21:9
Mental model:
zdx imagine -p "A visual explanation of Git branching using a subway map analogy. The main branch is the main line, feature branches are branch lines that split off and merge back. Commits are stations. Labeled with Git terms (main, feature, merge, rebase). Clean, colorful transit map style." --aspect 16:9
Resolution: Default is 1K. Use --size only when you need higher resolution.
Image editing (--source)
Use --source / -s to provide one or more input images for editing. The prompt describes what to change.
Single-image editing
Inpainting (add/modify elements):
zdx imagine -p "Add a small knitted wizard hat on the cat's head" -s cat.png
Inpainting (remove elements):
zdx imagine -p "Remove the person from the background, fill with natural scenery" -s photo.png
Style transfer:
zdx imagine -p "Transform this photograph into Van Gogh's Starry Night style. Preserve the composition but render with swirling impasto brushstrokes and deep blues and bright yellows." -s city.jpg
Outpainting / aspect change:
zdx imagine -p "Recreate this image as a cinematic ultrawide banner, extending the background naturally" -s hero.png --aspect 21:9
Color/mood adjustment:
zdx imagine -p "Make this scene a warm golden hour sunset, keeping all subjects the same" -s photo.jpg
Multi-image composition
Provide multiple --source flags to combine images into a new scene:
Group composition:
zdx imagine -p "An office group photo of these people, making funny faces" \
-s person1.png -s person2.png -s person3.png --aspect 5:4
Product mockup:
zdx imagine -p "Create a professional e-commerce fashion photo of this model wearing this dress" \
-s model.jpg -s dress.jpg
Composite with reference:
zdx imagine -p "Place this logo in the bottom-right corner of this banner image" \
-s banner.png -s logo.png
Editing prompt tips
Be specific about what to change — "change only the blue sofa to brown leather" works better than "change the sofa".
Describe what to preserve — "keep the rest of the room exactly the same, preserving style and lighting".
Don't reconstruct the original — use direct editing instructions, not a description of the whole image.
For detail preservation — describe critical details (face, logo) in the prompt alongside the edit instruction.
Supported formats: PNG, JPEG, GIF, WebP. Total request must be under 20MB (all source images combined).
Other prompt styles
For non-explainer use cases, see references/prompt-templates.md for templates and examples covering:
Photorealistic scenes
Stickers & illustrations
Text & typography in images
Product mockups
Minimalist / negative space
Cinematic landscapes
Logos
General prompting tips
Be specific about the subject — name the subject, action, environment, context.
Declare the style explicitly — without a style keyword the model guesses.
Use narrative descriptions — paint a picture with words, not keyword lists.
Iterate quickly — small wording changes produce big differences. Tweak rather than rewrite.
Workflow tips
Always view the generated image after creation to verify quality.
Images are saved to $ZDX_ARTIFACT_DIR when set. Use --out for descriptive filenames; the path is written exactly as given.
If the model returns no images, check the prompt for policy violations or try rephrasing.
Use --size 2K or 4K only when you specifically need higher resolution.
For iterative refinement, keep the base prompt and tweak one element at a time.
For editing, use --source with a direct edit instruction — don't describe the whole image.
Multi-image composition works best with 2–5 source images. More images = more ambiguity.
