Reasoning IMAGE STUDIO GUIDE
← Back to Image Studio

Image Studio — How to Use

A canvas-centred AI image generation workspace. This guide explains every control, shortcut, and workflow so you can create professional images without guessing.

🚀
Quick start — make your first image in 30 seconds
1
Type a prompt in the floating bar at the bottom of the screen. Describe what you want to see — subject, setting, mood, style. Example: "A golden retriever sitting in a sunlit café, warm bokeh background, photorealistic"
2
Press the Generate button (or / Ctrl). A progress bar appears on the canvas while the image is being generated.
3
Your image appears on the canvas. From here you can download it, right-click for more options, generate variations, or switch to inpaint mode to edit specific areas.
Not sure what to write? Click the AI prompt writer button in the bottom bar. It will automatically generate a high-quality, detailed prompt for you — or expand whatever you've already typed into a more complete description.

🎬
Prompt bar (the floating bar at the bottom)

The prompt bar floats above the status bar, spanning the width of the canvas. The 🎬 clapperboard icon on the left is a visual reminder that your words are the "script" the model follows.

ElementWhat it does
Text area Type your image description here. The bar expands as you type. Maximum 1000 characters — the counter below shows how many you've used.
Character counter Shows used / 1000. Keep prompts under 800 characters for best results with DALL-E 3.
✦ AI prompt writer See the next section for full details.
Generate button Submits your prompt with all current settings (model, style, ratio, etc.) and starts generation.
Tip: The style chips, lighting pills, and color mood pills in the right panel automatically append technical keywords to your prompt behind the scenes — you don't need to write "cinematic film grain" yourself if you've selected the Cinematic style chip.

AI prompt writer

The AI prompt writer button solves the hardest part of image generation: knowing what to write. Most people know roughly what they want but don't know how to phrase it in a way the model understands.

📝 Empty prompt box
Click the button with nothing in the text area. The AI will write a complete, professional prompt for a high-quality image from scratch — covering subject, lighting, composition, mood, and technical quality keywords.
💡 With a seed idea
Type a rough idea first (e.g. "coffee shop at night"), then click the button. The AI will expand your idea into a detailed, polished prompt while keeping your original concept.

The prompt is streamed character by character into the text area so you can watch it being written in real time. The button is disabled while streaming to prevent duplicate calls.

💡
After the AI writes your prompt, you can still edit it before generating. The AI's output is a starting point, not a final answer.

⚙️
Left panel — generation settings
ControlWhat it doesRecommended
Model DALL·E 3 produces higher quality, more prompt-accurate images. DALL·E 2 is faster and cheaper but lower quality. DALL·E 3 for most uses
Aspect Ratio Sets the image dimensions. 1:1 is square (best for product shots), 16:9 is landscape (best for scenes, banners), 9:16 is portrait (best for phone wallpapers, social stories). 1:1 for products, 16:9 for scenes
Quality Standard is faster and uses fewer credits. HD adds a second pass for finer detail and sharper textures — especially noticeable in faces, fabric, and complex backgrounds. Only available on DALL·E 3. HD for final exports
Inference Steps How many refinement passes the model runs. Higher = more detail but slower. Range: 10–100. Has no effect on DALL·E 3/2 (these are cloud models with fixed step counts) — useful when local models are added. 30 (default)
CFG Scale How strictly the model follows your prompt. Low values give more creative freedom; high values are more literal. Range: 1–20. 7–8 for balanced results
Seed A number that determines the random starting point for generation. Using the same seed with the same prompt will produce very similar images, useful for iterating on a specific composition. Click 🎲 for a random seed, or leave blank for fully random. Blank for variety, fixed for iteration
Batch How many images to generate at once. 1 = single image. 2 or 4 = multiple images loaded into the variation strip below the canvas so you can compare and pick the best one. Uses credits proportionally. 1 for speed, 4 for exploring options
Aspect ratio note: DALL·E 3 only supports 1024×1024, 1792×1024, and 1024×1792. Ratios like 4:3 and 2:3 are automatically mapped to the closest supported size. The status bar shows the actual pixel dimensions that will be used.

🎨
Style presets (right panel)

The six style chips translate a visual aesthetic into technical prompt keywords automatically. Click a chip to apply it — the active chip has an amber border. The selected style's technical keywords are appended to your prompt at generation time; you don't need to write them yourself.

PHOTOREALISTIC
CINEMATIC
ILLUSTRATION
OIL PAINTING
MINIMAL
SURREALIST
StyleTechnical keywords addedBest for
Photorealistic (none — your prompt is used as-is) Product photos, portraits, nature, architecture
Cinematic cinematic film still, anamorphic lens, film grain, color graded, letterbox Dramatic scenes, storytelling, moody landscapes
Illustration digital illustration, flat design, vibrant colors, clean linework, vector art Character art, children's books, app icons, social media
Oil Painting oil painting on canvas, impasto texture, expressive brushstrokes, chiaroscuro Fine-art style portraits, still life, landscapes
Minimal minimalist composition, generous negative space, clean geometry, Swiss graphic design Brand imagery, editorial, logo concepts, covers
Surrealist surrealist dreamscape, impossible geometry, melting forms, hyperrealistic fantasy Album covers, concept art, fantasy, abstract imagery

💡
Lighting & Color mood (right panel)

These two rows of pills fine-tune the atmosphere of your image independently of the style preset. Only one pill per row can be active at a time.

☀️ Lighting
Natural — outdoor ambient light, no extra keywords
Studio — softbox, professional product photography look
Dramatic — deep shadows, rim lighting, chiaroscuro contrast
Neon — cyberpunk glow, fluorescent reflections
Golden hr — warm backlight, sun flare, bokeh highlights
🎨 Color mood
Neutral — no colour grading added
Warm — amber and golden tones
Cool — blue-silver, desaturated feel
Vivid — high saturation, punchy contrast
Muted — film-fade, soft desaturated palette
Mono — black and white, high-contrast monochrome
💡
Combine style + lighting + mood for precise control. Example: Cinematic style + Dramatic lighting + Warm mood produces a high-contrast orange-tinted film still — without writing any of those technical terms yourself.

🚫
Negative prompt

The negative prompt field (bottom of the right panel) lets you tell the model what not to include. Anything you type here will be avoided.

Common things to add:
blurry, low quality, watermark, text, signature, distorted, extra fingers, bad anatomy, ugly, cropped, out of frame, duplicate, overexposed, underexposed

Note: DALL·E 3 and 2 do not natively support negative prompts via the API. The text is currently stored for future model support. For now, adding the negative concept to your main prompt works best: e.g. write "a clear sharp photo, no text or watermarks".

🖼
Canvas & hover tools

The canvas is the centre of the studio. Before generation it shows a placeholder. After generation your image fills the canvas and is the sole focus of the interface.

Hover over the canvas to reveal four tools in the top-right corner:

IconAction
✏️ InpaintToggle inpaint editing mode (see below)
⤢ FullscreenOpen the image full-screen in a new tab
⬇ DownloadDownload the current image as PNG
📤 ExportOpen the export presets panel (platform-sized exports)

🖌
Inpaint mode — edit specific areas

Inpaint mode lets you paint over an area of the generated image and regenerate only that region — leaving the rest of the image intact.

1
Generate an image first. Inpaint requires an existing image on the canvas.
2
Click the ✏️ button (top-right of canvas, or from the right-click menu). A floating toolbar appears at the top of the canvas.
3
Select a tool — Brush 🖌 for freehand painting, Lasso ⬡ for polygon selection, Auto-select ✦ for smart area detection, or Eraser ⬜ to remove mask paint. Use the slider next to the tools to change brush size.
4
Paint over the area you want to change. Then update your prompt in the prompt bar to describe what should appear in that area.
5
Click Apply to regenerate the masked region. Click ✕ to exit inpaint mode without applying.

🔢
Variation strip

The strip below the canvas holds up to 4 variation thumbnails (V1–V4). When you generate with Batch set to 2 or 4, all images are placed here side by side.

ActionHow
Switch between variations Click any thumbnail — the selected image appears on the main canvas
Generate variations of current image Click the Variations button on the right of the strip. This re-runs the current prompt with Batch=4 to produce 4 alternatives
Export a specific variation Click its thumbnail (it becomes the active canvas image), then use Download or the Export panel
💡
Set Batch to 4 before generating when you're exploring a concept. Set Batch to 1 when you have a refined prompt and want the best single image.

🖱
Right-click menu

Right-click anywhere on a generated image to open a context menu with every available action in one place.

Menu itemWhat it does
Download HDSave the full-resolution image as PNG (⌘S)
Copy to clipboardCopy the image to your clipboard for pasting into other apps (⌘C)
Share linkCopy the image URL, or use the native share sheet on mobile
Open fullscreenView the image in a new browser tab at full size (F)
Edit (inpaint)Enter inpaint mode to paint and regenerate specific areas
Generate more like thisRe-run the current prompt with the same settings to get another version
Save to libraryAdds the image to the History panel in the right sidebar
DeleteRemove the image from the canvas and clear the variation strip (Del)

📤
Export & download

Click the 📤 Export button (canvas hover tools, top-right) or the Export ↗ button in the variation strip to open the export panel.

Platform presets — one click to resize and download at the exact dimensions required by each platform:

PresetDimensionsUsed for
🛍 Shopify Product800 × 800 pxShopify product listings
📸 Instagram Post1080 × 1080 pxInstagram feed posts
📦 Amazon Main2000 × 2000 pxAmazon main product image
🐦 Twitter Card1200 × 675 pxTwitter/X link cards and headers
💼 LinkedIn Banner1584 × 396 pxLinkedIn company/personal banner

Format buttons — export the image at its original size in your chosen format:

FormatBest for
PNGMaximum quality, no compression — best for print or further editing
JPGSmaller file size — best for web pages, email, social media
WebPModern web format — best quality-to-size ratio for websites
How platform export works: The browser draws the image onto a hidden canvas at the target size and triggers a download. No data is sent to the server.

🕘
History

The History grid at the bottom of the right panel stores your last 12 generated images in your browser's local storage — they persist between sessions without needing an account.

ActionHow
View a past imageClick any thumbnail — it reloads onto the canvas and restores the prompt
Save manuallyRight-click the canvas → Save to library
Clear historyClick the Clear link next to the History label
History is stored in your browser only. Clearing browser data or using a different browser will erase it. For permanent storage, use the Download or Export options to save files to your computer.

⌨️
Keyboard shortcuts
ShortcutAction
⌘ ↵ / Ctrl ↵Generate image (works from inside the prompt bar)
⌘ S / Ctrl SDownload the current image as PNG
FOpen image fullscreen in new tab
DelDelete the current image from the canvas
EscClose any open modal, menu, or panel
Right-click on imageOpen context menu with all actions

Shortcuts work when the prompt bar is not focused (except ⌘↵/Ctrl↵ which works from inside the prompt bar).


📊
Status bar

The thin bar at the very bottom of the screen shows a live summary of all active settings so you always know exactly what will be generated before you click Generate.

ItemMeaning
● ReadyGreen dot = system is idle and ready. Turns into a spinner during generation.
ModelThe currently selected AI model (e.g. DALL·E 3)
SizeThe actual pixel dimensions that will be sent to the API (after ratio mapping)
StyleThe currently active style preset
QualityStandard or HD
Est. costApproximate credits this generation will use, based on model + quality + batch size
CreditsYour current credit balance

💰
Credits & generation cost
Setting combinationEstimated credits
DALL·E 3 · Standard · 1024×1024 · Batch 1~2 credits
DALL·E 3 · HD · 1024×1024 · Batch 1~4 credits
DALL·E 3 · Standard · 1792×1024 or 1024×1792 · Batch 1~3 credits
DALL·E 3 · HD · non-square · Batch 1~5 credits
Any setting · Batch 44× the single-image cost

Credits are deducted from your account balance after each successful generation. If your balance reaches zero, new generations are blocked until credits are added. Sign in to persist your credit balance across devices.


FAQ
Why did generation fail?
The most common causes are: no OPENAI_API_KEY configured on the server, insufficient OpenAI billing, or a content policy violation in your prompt. The error message from the server will indicate which one.
Why does the image look different from my prompt?
DALL·E 3 adds its own creative interpretation. Try being more specific: add camera angle, subject position, lighting direction, and background details. Use the AI prompt writer to expand a vague idea into a detailed prompt.
Can I use 4:3 or 2:3 ratios?
Yes — these ratios are available in the left panel and are automatically mapped to the nearest DALL·E-supported size (1792×1024 for landscape, 1024×1792 for portrait). The exact output size is shown in the status bar.
Does the AI prompt writer use credits?
Yes — the AI writer calls the main language model to compose the prompt, which deducts a small number of credits (the same as a general chat query, not a full image generation).
Why does "Copy to clipboard" fail?
Clipboard access requires HTTPS or localhost in most browsers, and the browser may ask for permission the first time. If the copy fails, use Download instead and paste the file manually.
Can I generate NSFW content?
No. DALL·E 3 enforces OpenAI's content policy. Prompts that violate the policy will be rejected with a 502 error. The system does not attempt to bypass these restrictions.

Ready to generate?

Open the Image Studio and start creating.

← Back to Image Studio