Create once.
Generate forever.
Vellux trains a private likeness model for each of your characters, then generates on-brand images and video in one canvas — text-to-image, inpainting, image-to-video, and an AI agent that handles the busywork.
Create your account
Start generating in minutes.
A full pipeline, built for consistency.
Four connected tools that keep every character on-model, from first train to final export.
Characters
Train a private likeness LoRA per character — one per base model, on Z-Image and Krea 2. Lock in a face that comes back consistent across every generation.
Studio
Text-to-image, inpainting, and image-to-video or video-to-video in one canvas. Mask a region to regenerate just that part, and pick any model from a single picker.
Agent
A chat agent that knows your characters and references, writes prompts, and runs generations — plus a visual builder for scheduled, multi-step automations.
Gallery
AI semantic search finds any shot by describing it in plain words. Recreate a result with the exact same settings, or send it straight to any generator.
From reference to release in three steps.
Build a character
Upload a few reference photos. Vellux trains a private likeness model — no prompt engineering required.
Generate on brand
Describe the shot in Studio — image or video — or let the Agent do it. Every result stays true to your character.
Batch and organize
Queue a full set in one run, export what you need, and keep everything searchable in your Gallery.
The same face, every generation.
Your next drop starts here.
Create your account and put your characters to work — generate on-brand images and video in minutes, and own every model and asset.