Skip to main content
All terms
Glossary

Identity lock

Definition — Fixing a persona's facial identity at training time so that every subsequent generation shares the same face and build. The alternative — re-describing a character in each prompt — produces visible identity drift between images.

There are two places an identity can live: in the prompt or in the model. Prompt-level identity ('a woman in her late 20s with auburn hair...') is re-interpreted from scratch on every generation, so each image lands somewhere slightly different in face-space. Identity lock moves the face into the model itself at training time, so generations vary in everything except who is depicted.

Locking the identity does not lock the content. Scene, wardrobe, lighting, mood and framing all remain prompt-controlled — and on InfluencerForge.app they are typically set per shoot with Forge Styles, so the persona stays constant while the look changes. The first photoshoot guide shows the split in practice.

It is a strong guarantee, not an absolute one. Extreme stylization prompts or contradictory descriptors can still pull a likeness off-center. When that happens the fixes are procedural — tighter situational prompts, seed reuse, and in stubborn cases a retrain with a broader reference set. The persona drift guide and consistency guide cover the toolkit.

Frequently asked questions

From theory to practice

See it live with 100 free credits

No card required. Generate with showcase models on the free tier, then train your own persona on any paid plan.