Retouch Backgrounds and Objects
Start the Lesson
By the end you can plan a retouch on paper before the software opens; run an AI background extraction as a directed first pass, bring it into GIMP as a layer mask over the original, and name at full zoom where it failed; clean a difficult edge by hand on the mask so the original pixels survive; vanish an object with clone and heal on its own repair layer so a person who knows the photo cannot find it; export each product with transparency and on white, paired with its original under one stem; and keep the receipts and the label on every piece.
Submit Your Work
What You Turn In
- Three before-and-after pairs, with the transparent PNGs
- Each product on clean white with its shadow; clean at arm's length, no halo at full zoom; one stem per product in the MD folder
- One layered working file
- Every cut is a layer mask, the original pixels underneath; the repair on its own layer, the object back when you hide it
- One difficult edge by hand
- Hair, fur, mesh, frayed cloth or glass, inside your pairs — no halo at full zoom
- One object removal
- A neighbour who saw the original hunted for the repair and did not find it
- Receipts, one AI critique and the method note fill in on the hub
- Spec before each first pass, where it failed, what your hand repaired; the critique checked on your own pixels; the piece labelled AI-assisted — directed by me
- Daily post
- One pair to the track channel with its label; one row in the posting log
- Clip of the Day: On White, for the Catalogue fill in on the hub
- 15–20 seconds, 9:16: each product from its shelf to white with its shadow; captions burned in; the end card with your mark and the label; a 720p MP4 under 30 MB in your Mufasa folder, its Google Drive share link and the receipt in Asha
How
- Asha, by the week 4 Saturday update.
Done when
- Three before-and-after pairs + transparent PNGs: Clean white with shadows; clean at arm's length, no halo at full zoom; exported transparent and on white
- Layered working file: Every cut a layer mask, original pixels survive; repair on its own layer
- One difficult edge by hand: Hair, fur, mesh, cloth or glass: clean at arm's length, no halo at full zoom
- One object removal: On its own layer; a neighbour who saw the original could not find it
- Receipts + AI critique + method note: Spec before each first pass, where it failed, what the hand repaired; critique verified on own pixels; piece labelled
- Daily post: One pair posted with its label; one log row
- Clip of the Day: On White, for the Catalogue: 3 stars: as 2, plus the removal beat (the object gone, the repair not found by a partner) and the hook in the first two seconds · 2 stars: 15–20 s; 9:16 at 720p, MP4 under 30 MB; captions burned in inside the safe area; end card with the trainee's mark and the label; receipt filed; three products from shelf to white with shadow; labelled AI-assisted — directed by me · 1 star: not 9:16 or outside 15–60 s, a halo visible on a phone, or no receipt