How to create an AI photo with a parent who has passed away
A step-by-step guide to creating a meaningful AI photograph of yourself with a deceased mother or father. Reference photo selection, scene choice, and the emotional considerations worth thinking through first.
By Jiuhong Deng · · Updated
Most people who land on a guide like this share roughly the same situation: a parent died, milestones have happened since, and the photographs from those milestones feel incomplete. This is the most common reason people use AI memorial photo tools in 2026 — and the use case Lover Snap was built around.
The seven-step workflow
1. Decide which era you want to remember them at
The single biggest decision is what age and era you want your parent to appear in the generated photo. Most people pick the version of their parent they remember most clearly — often a 5-10 year span. If you mix photos from when they were 30 and when they were 70, the AI will struggle to settle on a face, and the generated character will look generic.
2. Gather reference photos
Aim for 3-16 photos from your chosen era. Mix close-ups (face and shoulders), medium shots, and at least 2-3 full-body shots. Different settings and lighting helps. Avoid sunglasses, hats, group photos where they’re small in frame, and heavily filtered images.
If the photos are physical prints, scan them at 300-600 DPI. Many home scanners default to 200, which is too low. iPhone photos in HEIC format need conversion to JPG — easiest path is to use the Photos app’s “Save Image” option which auto-converts.
3. Train the AI character
In Lover Snap, tap Create a character, select your reference photos, and start training. Expect about 30 minutes. You’ll get a notification when training is complete. Close the app while it runs.
4. Train an AI character of yourself
If you want to be in the photo, train your own character too. Use photos of yourself from around the age you want to be in the scene. If you’re 40 today and want a current-day family photo, use photos from the last 1-2 years.
5. Pick a scene
Some scene ideas that work well for parent memorial photos:
- A walk in the neighborhood you grew up in
- The kitchen at breakfast, soft window light
- A wedding scene (yours or a sibling’s)
- Holiday dinner at the family table
- Reading together in a quiet living room
Lover Snap has scene presets for many of these, plus a custom prompt field for specifics.
6. Generate in burst mode
Generate 16 photos at once. AI image generation is variable — about 75% of generations are good, 10% are exceptional, and the rest have artifacts. Generating 16 gives you a real chance of getting at least one or two photos that feel true.
7. Pick the keepers, delete the rest
This is harder than it sounds. Some photos will feel right immediately; others will feel wrong in ways you can’t quite name. Trust your reaction. Deleted photos don’t haunt your account. Save the keepers, maybe print one — many families frame a single generated memorial photo for a quiet corner of the home.
What to plan for emotionally
The first generation session usually takes longer than expected. People who casually open the app to “try it real quick” sometimes find themselves sitting with the result for an hour. Plan for that.
If you have other family members who may have feelings about the project, consider talking to them before generating photos you intend to share publicly. Grief is plural, and what comforts one family member can hurt another.
Style suggestions
For memorial photos that feel like real family album photographs:
- Camera style: 35mm film (Kodak Portra style) or Polaroid
- Lighting: Golden hour outdoor, or soft window light indoor
- Canvas ratio: 3:2 (printable) or 4:5 (general portrait)
- Shooting angle: Eye-level or gentle hero angle
Avoid harsh midday sun, low-angle drama shots, and clinical studio lighting for memorial photos. They can feel emotionally cold.
What if I only have a few photos?
Three or four photos can produce a recognizable character. If you have access to family albums from relatives, scanning a few more makes a real difference. Single-photo training is possible but produces generic results — the AI can’t learn a face from one angle.