Skip to main content
All terms
Glossary

Model training

Definition — The one-time process of teaching an image-generation model a specific persona's identity from a set of reference photos. After training, the persona can be generated in unlimited new scenes without re-describing its appearance in each prompt.

In plain language: training shows the model the same person from many angles until it learns what makes that face that face — bone structure, proportions, skin, build — as an abstract identity rather than a set of memorized photos. That is why a trained persona can appear in scenes, outfits and lighting that exist in none of the reference images.

On InfluencerForge.app training is a one-time 720 credits per model; refreshing a model later with an expanded reference set is 800 credits. After training, the persona generates at normal image prices (12–20 credits per photo). The first-model Academy guide walks through the process end to end.

The single biggest lever on training quality is the reference set. Varied angles, lighting and expressions produce a robust identity; a stack of near-identical selfies produces a brittle one that drifts the moment a prompt leaves its comfort zone. See how to build a brand-safe reference set before you spend the credits.

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.