# Sketch-to-Image, Copy Design, Vectorize & Enhance

How to turn sketches into finished images, copy a design onto another image, trace line art into editable vectors, and refine or reverse-engineer prompts in Framo.

## Sketch to Image (Realistic variant)

Turns a hand-drawn sketch on the canvas into a realistic image, with no style reference.

To use it:
1) Select the sketch image on the canvas.
2) Open the AI Tools menu (the wand button on the node toolbar) → **Sketch Tools** → **Sketch to Image** ("Turn a sketch into a realistic image").
3) In the dialog, type a few words for what the sketch shows (e.g. "sports car, front view") to keep the render on subject.
4) Optionally tick **Clean up the sketch first** (default off; adds credits) — recommended only for rough linework, skip it if your lines are already clean.
5) Click **Generate**.

The result lands as a new image on the canvas (the dialog closes immediately and progress is shown on a pending tile). In the Pixi edit canvas it lands as a layer instead.

## Sketch to Image with Reference (Reference variant)

Same as above, but renders the sketch in the visual style of another image you pick.

To use it:
1) Select the sketch on the canvas, open AI Tools → **Sketch Tools** → **Sketch to Image with Reference** ("Render a sketch in a style you pick").
2) An eyedropper arms — click any image on the canvas to use it as the style reference (Esc cancels).
3) The same dialog opens (titled **Sketch to Image · Reference**): enter the subject, optionally tick **Clean up the sketch first**, click **Generate**.

The picked reference contributes style only; the sketch defines the composition. This variant is hidden inside the Pixi edit canvas (the on-canvas style pick can't run mid-edit).

## Clean Sketch

Tidies a rough sketch's linework into a clean, shaded line drawing — no realistic render. The output is often useful on its own.

To use it: select the sketch, open AI Tools → **Sketch Tools** → **Clean Sketch** ("Tidies the linework — no render"). Enter a few words for the subject and click **Generate**. One render charge.

This is the same clean pass that the two Sketch-to-Image variants run when you tick "Clean up the sketch first" — exposed here as a standalone tool.

## Copy Design

Transfers the design/styling of one image onto another (e.g. swap one car's body design onto your subject).

To use it:
1) Select the target image on the canvas.
2) Open AI Tools → **Copy Design** (eyedropper icon).
3) Pick an intensity from the submenu:
   - **Subtle — light design hint**
   - **Balanced — clear design swap**
   - **Bold — reference dominates**
4) An eyedropper arms — click the design-reference image on the canvas (Esc cancels).

The result generates onto the canvas. Higher intensity lets the reference dominate more. (This tool is hidden inside the Pixi edit canvas.)

## Vectorize Sketch

Traces a raster layer's line work into smooth, editable Bézier paths (a native path layer you can keep editing with the pen tool). This runs on a layer inside the [Pixi 2D editor](/docs/product/editor-2d).

To use it:
1) In the edit canvas, select a raster layer and trigger **Vectorize** from the layers panel.
2) In the **Vectorize Sketch** dialog, pick a trace mode:
   - **Centerline** — one stroked path per clean sketch line
   - **Ridge** — centerlines from shaded / tonal / faint art
   - **Outline** — closed loops around the ink boundary
3) Tune with the toggle and sliders, watching the live blue-path preview over the dimmed source:
   - **Clean up lines** (on by default) — merges tiny segments into long, smooth character lines; shows a **Connect gaps** slider for bridging fragmented strokes.
   - **Ink threshold** / **Ridge strength** — how much ink is captured (label depends on mode).
   - **Line scale (σ)** — ridge mode only.
   - **Simplify (ε)** — fewer, simpler anchors.
   - **Smoothing** — curve smoothness.
4) Click **Create Path Layer**. The trace lands exactly over the source layer; the footer shows path/anchor counts.

## Prompt Enhance (inline panel)

Rewrites the prompt in your prompt field using AI. Available wherever the prompt enhancer panel appears in the prompt input.

- With text already in the prompt, four style buttons appear: **Polish** (sharpen and improve), **Distill** (strip to the essence), **Enrich** (add technical depth), **Reimagine** (transform and elevate). Click one to rewrite the whole prompt in that style.
- The chat box below is always available: type an instruction (e.g. "make it nighttime") to modify the prompt, or — when the prompt is empty — type an idea and the AI writes a full prompt. The mic button dictates by voice.
- If you highlight a fragment of the prompt first, the chat box scopes its edit to only that fragment.

Each enhance/chat action costs 1 prompt-enhancement credit.

## Prompt Enhance (chat refinement)

A multi-turn chat for iterating on a generation prompt with the AI, opened from the chat bar. This is a **Pro plan** feature.

To use it:
1) In the chat bar's top row, click the **wand button** ("Refine prompt with AI"). (Free users get an upgrade dialog instead.)
2) Whatever you'd typed in the chat bar seeds the first message. Send follow-up instructions, or click a preset chip — **Polish / Distill / Enrich / Reimagine** — to refine. Each reply is a complete proposed prompt you can keep iterating on.
3) If you have reference images selected on the canvas, they show in the editable strip below the chat. Clicking a tile inserts an `@image1`, `@image2`, … mention at the cursor; typing `@image2` (or "image 2") in your instruction points the AI at that numbered reference.
4) Click **Use this prompt** on any reply to accept it and return to the chat bar. Closing with the **X** discards the chat.

A word meter (top-right ring) tracks the free per-conversation allowance; closing resets it. See also the chat bar's **Enhance** pill, which is a separate one-shot toggle that auto-analyzes the input image and reconciles the prompt at generation time. See [Chat bar: generating images](/docs/product/chat-image-gen).

## Reverse Prompt (Default)

Generates a detailed, natural-language prompt that could recreate a selected image — useful for understanding or reusing an image's look.

To use it: select one image on the canvas, open AI Tools → **Reverse Prompt** → **Default** (1 credit). It analyzes the image and produces a single flowing prompt covering subject, composition, camera/lens, lighting, color, environment, style, and quality. In the result dialog you can **Use Prompt** (loads it into the prompt field), **Regenerate**, or **Clear**. There's also a **Reverse Prompt** popover (image icon) that does the same and lets you edit the text before using it.

## Reverse Prompt (Structured)

Like Default, but returns a structured breakdown of the image (objects, lighting, aesthetics, photographic characteristics) instead of one paragraph.

To use it: select an image, open AI Tools → **Reverse Prompt** → **Structured** (2 credits). The result is shown as a structured analysis; **Use Short Description** loads its short description into the prompt field. Also available from the Reverse Prompt popover via the **Structured** toggle.

## Reverse Prompt for HDRI (panorama)

Inside the HDRI generator, this analyzes a chosen image into a short, simple scene description tuned for 360° panorama generation (one sentence, location + key elements + lighting; no camera terms).

To use it: in the **HDRI generator** dialog, choose to create from an image, select your source image, and let it analyze. The resulting one-line description fills the **Describe Scene** step, where you can edit it before generating the panorama.