Use Cases

How to Batch-Replace Objects in 50 Images: Select and Describe

Select the object to replace and enter the original and new (e.g. "red ceramic mug" → "blue glass vase"), and AI generates the new object inside the selection while keeping position, perspective, reflections and scene lighting consistent; pixels outside the selection stay original. Up to 50 images share one description.

Eyesme Team
Eyesme Team
Published on
How to Batch-Replace Objects in 50 Images: Select and Describe

The short answer: box the object to replace (required), type the original and the new object (required, e.g. "red ceramic mug" → "blue glass vase"), and AI generates the new object inside the selection; position, perspective, reflections and scene lighting stay consistent, and pixels outside the selection remain untouched. Up to 50 images can share one replacement description.

What object replacement does

It doesIt doesn't
Replace only the selected object; outside stays originalTouch anything outside the selection
Keep the original position, scale and perspectiveRestructure the scene or other objects
Blend the new object into scene lighting (reflections, shadows)Make you cut out, composite and relight by hand
Share one description across a 50-image batchSet different replacements per image

It is built for objects, furniture, food and props; people and animals can be attempted but identity preservation isn't guaranteed (see the boundary below).

How to write the descriptions accurately

The formula: a clear noun + material/color, no abstractions:

  • ✅ "red ceramic mug" → "blue glass vase";
  • ✅ "wooden chair" → "black leather modern lounge chair";
  • ❌ "that cup" → "something nicer".

Mirror-like, transparent or frosted materials go into the optional notes ("frosted glass material, keep the original shadow"); the model handles them more precisely when named.

How to replace step by step

  1. Upload images: one or many, up to 50; JPG / PNG / WebP, each up to 10 MB;
  2. Box the object: required; the tighter the selection hugs the object's edges, the more natural the replacement;
  3. Enter the original and the new object: both required (see above); add material/color notes if needed;
  4. Download: review each result, then download singles or the whole batch as a zip (files keep their names under the eyesme-replaced prefix). One description per batch, exactly right for product colorways and style variants.

Four scenes in practice

  • Product colorway: the red ceramic mug becomes a blue glass vase, with table, notebook and plant untouched:

Red enamel mug to blue glass cup before/after: original above, replaced belowRed enamel becomes blue glass; reflections and highlights recompute for the new material

  • Furniture swap: the wooden chair turns into a black leather modern chair, room perspective and floor shadow intact:

Wooden chair to leather chair before/after: original above, replaced belowThe style changed; the leg shadow and rug texture still connect seamlessly

  • Food swap: pasta becomes a sushi platter, with plate, table and cutlery unchanged:

Pasta to sushi before/after: pasta above, sushi belowSame plate, same light; it reads like one photo session

  • Prop swap: the red umbrella becomes a black walking umbrella, scale and lean angle kept:

Red umbrella to black umbrella before/after: original above, replaced belowThe umbrella's shape and wall lean stay; color and material follow the description

Can people and animals be replaced?

You can try, but set expectations: this tool is designed for objects, furniture, food and props. For people or animals use clear photos and very specific descriptions; identity preservation isn't guaranteed. If you want someone gone rather than replaced, AI object erasing is the better fit.

vs manual compositing and general AI editors

Manual compositing / editorsAI object replacer
WorkflowFind assets, cut out, composite, relightSelect the object + enter original/new
Position & perspectiveManual alignment; easy to skewOriginal position, scale and perspective kept
Lighting & reflectionsGrade and shadow by handNew object follows the scene's lighting
Other elementsEasy to disturb the backgroundEverything outside the selection stays
BatchOne image at a timeUp to 50 images, one description
General AI editorsAI object replacer
Getting startedRewrite prompts and tune parametersSelect the object + enter original/new
TargetingEasy to replace the wrong objectSelection targeting; only the chosen object changes
Material & lightingExtra description neededPerspective, reflections and scene lighting kept
Batch consistencyTune each imageOne replacement description for the batch

In one line: manual compositing wins on asset control, general AI on freedom, and the replacer wins on "box it, swap it, lighting doesn't break, batch consistent".

Frequently asked questions

How does object replacement work?

Select the object, enter the original and new object (plus material notes), and AI generates the new object inside the selection while keeping position, perspective, reflections and scene lighting consistent.

Is a selection required? How should I draw it?

Yes. The selection marks the object to replace; a tighter selection gives a more natural result, and pixels outside stay original.

How should I write the original and new object?

Clear nouns plus material and color, e.g. "red ceramic mug" → "blue glass cup". Avoid abstract descriptions.

Will it change other objects or the background?

No. Only the selection is replaced; everything outside stays from the original, and the new object blends into the scene lighting.

Do perspective, reflections and shadows match?

Position, scale, perspective, reflections and shadows are matched as closely as possible; mention mirror, transparent or complex materials in the notes.

Can I process a batch?

Yes, up to 50 images at once with the same replacement description, ideal for a product series or similar scenes.

Can I replace people or animals?

You can try, but the tool is designed for objects, furniture, food and props; identity preservation isn't guaranteed for people or animals.

Which formats and output sizes are supported?

JPG, PNG or WebP, up to 10 MB each; output size adapts to the input, and the result card shows actual dimensions.

Features and details reflect the tool as of November 12, 2026 and may evolve; the before/after examples are from the tool's real demo assets, with the cover composited from the original and replaced halves.


Ready to swap objects in the frame? Eyesme's AI Object Replacer takes a selection plus two short descriptions, for up to 50 images with lighting and perspective intact. Just want something gone? Use AI object erasing.

Related use cases