Can I train a custom model or LoRA on my own images?
No — custom model and LoRA training are not currently available. Nano Banana Pro reference editing has replaced that workflow and needs no training at all.
Fauxto Labs does not currently offer custom model training or LoRA generation. If you have read otherwise in an older guide, that information is out of date.
What replaced it
Reference-based editing got good enough that training became unnecessary for most work. Nano Banana Pro in the Image Editor is exceptional at holding a person’s likeness while putting them in a completely new pose, outfit or scene — from a single reference photo, with no training step, no waiting and no per-model setup.
How to get a consistent character now
| Goal | Approach | Credits |
|---|---|---|
| Best likeness across new poses and scenes | Image Editor with Nano Banana Pro | 15–30 per image |
| Good likeness at lower cost | Image Editor with Nano Banana 2 | 10 per image |
| A recurring social persona | Influencer Studio in The Lab | 15 per image at 1K |
| A batch of styled portraits from one face | Photo Packs | 10 per photo |
| Reusing the same character across tools | Save it as a reference set in Projects | Free |
For video
Reference-to-video does the equivalent job. Attach reference images or video and models like Seedance 2.x, Hailuo 03, the Kling O3 family or Gemini Omni Flash will match that subject in the generated clip.
Last reviewed against the live product in August 2026.