← All docs

Docs

Edit with annotations

Draw your intent directly on an image with notes, text, arrows, and People cutouts, then Generate — the AI reads the annotations and applies them as edits.

Annotations let you mark up an image right on the canvas — sticky notes, text boxes, arrows and shapes, and People cutouts — and the AI reads those marks as instructions when you generate.

When to use annotations

  • You want to point at a specific spot in the image and say what to change there.
  • Describing the edit in words is awkward, but showing it with an arrow or a note is easy.
  • You want to add a realistic person exactly where you place them.

Steps

  1. Select the image or layer you want to edit.
  2. Add your marks on top of it — a sticky note carrying an instruction, a text box, an arrow or shape pointing at the area, or a People cutout. (See the annotate tools in Editor & Canvas Objects for how to place each one.)
  3. Write a prompt if you want to — or leave it empty and let the annotations carry the intent.
  4. Click Generate.
  5. When your marks overlap the target, a small “Would you like to include them in the generation?” panel appears. Choose Yes, Include to bake the annotations into the render, or No, Skip to generate without them. The X cancels.

How it works

When you generate, INTERSTITIAL finds every annotation that overlaps your target and composites them onto the image the AI receives. It also passes a structured description of each mark — a note’s text, a shape and whether it has an arrowhead, a cutout’s position and size — so the model reads them as directions rather than as pixels to trace.

  • On an image or the scene, the include/skip panel lets you decide each time whether the overlapping marks count.
  • On a frame, overlapping content is folded in automatically — no panel appears.
  • Sticky notes and text boxes contribute their words. Arrows and shapes contribute direction and emphasis. A few clear marks read better than a cluttered pile.

People cutouts

A People cutout is a person-shaped silhouette you drop onto the image to reserve a spot for a figure.

  1. Open the annotate tools and choose People to open the cutout gallery.
  2. Pick a silhouette and place it on the canvas. Drag to move it; use the corner handles to size it — height sets how large the person reads in the scene.
  3. Generate. The AI treats the cutout’s position and size as a placement guide and renders a realistic person in that spot.

Place several cutouts to populate a space, and scale each one to sit correctly at its distance.

Combine with references

Annotations stack with reference images. Attach a reference to set the look — a material, a mood, a style — while your annotations say where and what to change. The generation reads both together. See Reference images & Are.na to add one.

Tips

  • Put the note where you mean it. Placement matters — a mark only counts when it overlaps the image, layer, or frame you’re generating on.
  • Fewer, clearer marks win. Long paragraphs and overlapping scribbles dilute the intent.
  • Size People cutouts for depth. A smaller silhouette reads as farther away; a taller one reads as close.
  • Empty prompt is fine. With annotations present, the AI can work from the marks alone.
  • Costs vary by the model that runs the render — check the in-app pricing screen and Plans & Credits for specifics.