Sketch to image

The Sketch to image mode (Sketch) turns a canvas drawing into a final AI-generated image.

This is the right workflow when you want to control composition and shapes with a drawing (even very simple) and then let the AI deliver the final render: realistic photo, illustration, 3D, concept art, etc.

In this mode the sketch is the “map” of the scene. The prompt defines style, materials, light, and details.

Use Sketch if:

• you have an idea and want to visualize it quickly;
• you want to control framing and composition (before aesthetics);
• you want to prototype poses, layouts, product shots, or scenes;
• you want to turn a draft into a coherent, high-quality image.

• If you want to start from a real photo: use Single image edit / Background edit / Enhance quality.
• If you want to generate from text without drawing: use Generate image from text.
• If you need to merge real photos: use Compose multiple images or Seamless Blend.

• A sketch (even minimal) created in the editor.
• A prompt:
  • optional, but strongly recommended if you do not want default photorealism;
  • useful to specify style, light, colors, details.

When you start rendering, the app sends to the model:

• the sketch image (canvas);
• the prompt (if present);
• any style hints (when applicable).

If you do not provide a prompt, the model tends to render the sketch as a highly realistic photo.

In this mode you typically see:

• Mode: workflow selector (choose Sketch to image).
• Canvas/Sketch preview: shows the current sketch.
• Open sketch editor: opens the editor to draw.
• Clear: clears the sketch.
• Prompt: describes what the sketch should become (style and details).
• Render: generates the final image from the sketch.
• Generate variants (2-4): alternatives of the same render.
• Use as input: bring the result into Edit for markup refinements.

1. Draw a simple but readable sketch (silhouette + main elements).
2. Write a prompt describing:
   • subject,
   • environment,
   • style (realistic/illustration/3D),
   • light and palette.
3. Render.
4. Generate 2-4 variants and pick the best.
5. If needed, “Use as input” and refine in Single image edit.

In the “Mode” selector, choose Sketch to image.

Click Open sketch editor and draw on the canvas.

Practical tips for a sketch that works:

• draw big shapes before details (silhouette, main objects);
• use clear, readable lines;
• separate subject and background (even with simple outlines);
• if you want a specific pose, sketch it recognizably.

You do not need to be good at drawing: you need to communicate structure.

The prompt turns the sketch into a concrete image.

Example (photorealistic): “Realistic photo portrait in studio, soft softbox light, natural skin, 85mm, neutral gray background, high definition.”

Example (illustration): “Watercolor illustration, pastel palette, soft lines, dreamy atmosphere, minimal background.”

Click Render and evaluate:

• consistency with the sketch composition;
• light and style quality;
• details (hands, text, patterns).

• Use Generate variants (2-4) for quick alternatives.
• If you want to change style, adjust the prompt (often 1-2 lines are enough).
• If the output is close but needs local precision: Use as input -> Single image edit.

1. Main subject
2. Environment
3. Light
4. Style
5. Details/materials
6. What to avoid (negative)

“Transform the sketch into: [SUBJECT] in [ENVIRONMENT]. Light: [LIGHT]. Style: [STYLE]. Details: [DETAILS]. Avoid: text, watermark, artifacts.”

“3D render of [SUBJECT], realistic materials, studio lighting, soft shadows, high definition, clean look.”

“Concept art of [SCENE], cinematic atmosphere, light fog, palette [COLORS], coherent details, style [ARTIST/GENRE].”

• Without a prompt, the default tends to be photorealistic: use the prompt to change style.
• The sketch guides composition, but does not guarantee pixel-perfect accuracy: for refinements use Edit.
• Perfect text is not guaranteed: if you need precise text, add it later.

• Make the sketch more readable (clearer shapes, less ambiguity).
• Describe composition in the prompt (“centered subject, simple background”).
• Generate variants and pick the most coherent.

• Put the style at the start of the prompt (“style: …”).
• Repeat 1-2 keywords (e.g. “watercolor”, “photorealistic”).

• Add negatives (“avoid deformed hands”).
• Use “Use as input” and fix in Edit with markup.