Treat each person as a separate identity review, even when the finished cards belong to one collection. Use a clearly matched authorized source, generate and approve that person's artwork, then apply a shared design system to the set. A single instruction to make several similar-looking people is not a dependable way to preserve individual faces.
Create a source-to-card register
Give each subject a neutral internal label and record which reference belongs to them. Keep those labels out of public artwork unless they are intended names. Do not infer identity from appearance or mix files because two outfits look similar.
The register can be simple: person label, source filename, selected format, accepted output and approval status. Its purpose is to prevent operational mistakes as well as generation errors. A correctly rendered portrait paired with another member's name is still an incorrect card.
Share the design, not the face description
Define a few collection-level decisions: border treatment, typography area, contrast range and edition layout. Keep each person's pose and source separate. This lets the collection feel connected without forcing every face into the same smile, hairstyle or pose.
Collection direction: a series of vertical rehearsal-room cards with warm side light, a narrow cream border and a clear lower name area. For this card, use only the attached authorized person's identity. Give this person a relaxed standing pose with hands loosely behind the back. Other members will be produced and reviewed in separate tasks.
The example does not promise automatic multi-reference separation or batch identity locking. Use the input behavior available in Imagild Fan Merch. If a tool accepts several files, confirm whether it treats them as separate jobs or combined references before relying on it.
Review each card before comparing the collection
Compare a candidate with its own source first, at equal face size. Check facial proportions, hairline, expression and distinctive visible details. Do not approve a candidate merely because it resembles the aesthetic of the other cards.
After individual approval, view the set together. Check visual hierarchy, subject scale and background balance. A consistent set does not require every background to be identical; variation can be intentional as long as the collection-level decisions remain clear.
| Review stage | Main question | Typical failure |
|---|---|---|
| Identity | Is this the correct person's portrayal? | Face drift or source mix-up |
| Labeling | Does the text match this card? | Swapped name or edition number |
| Collection | Do the cards belong together visually? | One oversized face or unreadable background |
| Delivery | Are the approved files the exported files? | An older candidate accidentally included |
Keep a final contact sheet
A contact sheet with filenames makes a last review easier than opening files from memory. Mark rejected versions clearly in your local records and export only the approved selection. This is a manual quality step, not a claim that the website supplies a bulk approval dashboard.
Can I generate a group portrait instead?
You can plan one, but several interacting faces create a different review problem. Do not assume it will produce reliable individual card portraits automatically.
Does a shared reference improve consistency?
A shared visual brief can align design. Sharing the wrong person's image as an identity anchor can instead contaminate the result. Keep those roles distinct.