<![CDATA[Act2Know Notes]]>https://act2know.hashnode.devRSS for NodeWed, 28 Oct 2026 17:09:36 GMT60<![CDATA[Removing Tourists From Travel Photos With a Text Prompt (Quick Guide)]]>https://act2know.hashnode.dev/removing-tourists-from-travel-photos-with-a-text-prompt-quick-guidehttps://act2know.hashnode.dev/removing-tourists-from-travel-photos-with-a-text-prompt-quick-guideFri, 23 Oct 2026 03:28:21 GMTA tour group in front of the landmark no longer means an evening with the clone stamp. Image-to-image AI editors can remove people and objects from a plain-language instruction. The short version of the workflow:

  1. Shoot with removal in mind. Take a burst so the crowd covers less of the scene, leave space between your subject and strangers, and edit the full-resolution file.
  2. Write a three-part prompt. Say what to remove ("the tourists behind me, the trash bin"), what to keep unchanged ("the person in the teal coat: face, pose, clothes"), and what to fill with ("the same cobblestones and arcade"). Point at your subject by clothing color, not "me".
  3. Iterate in small steps. Remove the crowd first, then objects, then leftover shadows, checking subject edges and repeating patterns at 100% after each pass.
  4. Compare the subject before exporting and keep the original.
Three-panel illustration: the original crowded square with a barrier and trash bin, then pass 1 with the crowd removed, then pass 2 with the barrier and bin removed, the traveler unchanged throughout.

Removing the crowd first and the objects second gives the model one job at a time. (Illustration, fictional scene)

It handles background people and small clutter well. People overlapping your subject, heavily hidden landmarks, text and reflections are still hard, and the rebuilt background is a guess, so label edited photos honestly.

Full step-by-step version with all the prompts and examples: Remove Tourists From Travel Photos Without Photoshop: A Prompt-Based Workflow

(Disclosure: I build NanoPic, the image-to-image editor I used for the examples.)

]]><![CDATA[Restoring Old Family Photos With a One-Sentence Prompt (Quick Guide)]]>https://act2know.hashnode.dev/restoring-old-family-photos-with-a-one-sentence-prompt-quick-guidehttps://act2know.hashnode.dev/restoring-old-family-photos-with-a-one-sentence-prompt-quick-guideThu, 22 Oct 2026 03:34:47 GMTFaded, creased and torn family prints no longer need hours with a clone stamp. Image-to-image AI editors can repair them from a plain-language instruction. The short version of the workflow:

  1. Scan well. 600 dpi on a flatbed, or a phone photo in daylight with no flash, cropped to the print.

  2. Repair first, colorize second. Ask for the scratches, creases, dust and torn corner to be removed, and say explicitly to keep every face, expression and outfit unchanged. Then colorize the repaired image in a second pass, naming any colors you actually know.

  3. Refine in small steps. "Bring back natural skin texture", "make the colors less saturated, like a 1960s print", "only sharpen the faces".

  4. Export at full resolution and keep the original scan.

Side-by-side illustration: scratched black-and-white photo of a fictional grandmother and boy on a bench, and the same scene repaired and colorized.

It handles creases, scratches, fading and color casts well; faces that are mostly missing are still a guess, so say so when sharing.

I wrote up the full step-by-step version with all the prompts here: How to Restore and Colorize an Old Family Photo With One Sentence

(Disclosure: I build NanoPic, the image-to-image editor I used for the examples.)

]]>