How it works
Nine guided steps, grouped into three: build the character, train the model, then create.
Lock a face
Describe a look, generate a few options, and set your Hero. Every later step keeps that exact face.
Build & train
The studio walks you through defining the character, generating a dataset, and training it into a LoRA — guided the whole way.
Create forever
Test, then export. Now generate unlimited consistent photos of your character — the same face, every time.
Give them an identity
Describe your character in plain words — face, hair, body, and overall feel. The studio turns it into a fixed signature that every photo follows, so the look never drifts.
- Free description in any language → structured signature
- Head / identity and body, locked separately
- Reuse the same identity across every shoot
Stage 2 — Define IP
Pin down who they are — identity and body — so the look stays consistent in every photo.
Generate a whole dataset
Make lots of consistent photos across poses, outfits and scenes — then collect the keepers. That curated set becomes the training data for your character's model.
- Recipe-driven batches — one variation per line
- Collect the keepers, dismiss the rest
- Refine and auto-tag into a clean dataset
Stage 3 — Drafts (recipe-driven)
Generate lots of photos of your character across poses, outfits and scenes. Collect the keepers for the dataset.
cafe, candid
outdoors, golden hour
Train a model you own
Turn the dataset into a LoRA that reproduces your character on command — trained locally on your own GPU, uncensored, and yours to keep and reuse forever.
- One trigger word brings your character back
- Trains locally — no uploads, no filters
- Commercial-friendly models — you own the LoRA
Stage 7 — Train LoRA
This teaches the app your character’s face from all the photos. Takes a while — watch the progress and the sample shots.
Made for people who build characters
If you need a character that looks the same in every image — and you want to own it — this is for you.
Virtual influencers
One consistent face for a character account you grow over time.
AI model studios
Create multiple characters you own, each with its own LoRA.
Adult creators
Unfiltered, fully local — your content never leaves your PC.
Brands & mascots
A repeatable character face for campaigns and product shots.
Why local & one-time
You buy it once and run it on your own GPU. Compare that to a per-image AI service.
Inside the app
An illustrative look at the full 9-step pipeline.
Stage 1 — Create Hero (lock the face)
Make your character’s face. Describe the look you want, generate a few to choose from, and pick the best one — every step after this keeps that face.
Stage 2 — Define IP
Pin down who they are — identity and body — so the look stays consistent in every photo.
Stage 3 — Drafts (recipe-driven)
Generate lots of photos of your character across poses, outfits and scenes. Collect the keepers for the dataset.
cafe, candid
outdoors, golden hour
Stage 7 — Train LoRA
This teaches the app your character’s face from all the photos. Takes a while — watch the progress and the sample shots.
Stage 9 — Your finished assets
Everything this character left on your disk: the LoRA you use, leftover files, and the training photos.
Everything in the box
One desktop app, no add-ons.
Face lock
Pick one Hero face — every later step keeps it, exactly.
9-step pipeline
Guided from a face to a finished, trained model.
Your own LoRA
Train a model that brings your character back on command.
Unfiltered
No content filters — it’s your creative call.
100% local
Runs on your own GPU. Nothing is ever uploaded.
You own it
Commercial-friendly models; keep the LoRA & dataset.
Technical specs
What it runs on and what’s inside — all local, no accounts.
Windows 11 desktop app
NVIDIA GPU, 16 GB (e.g. RTX 4060 Ti / 5060 Ti)
Fully offline — no account, no API
Qwen-Image-Edit — face-lock editing
CHROMA — base image generation
Commercial-friendly licenses — sell what you make
9-step guided pipeline, one IP at a time
Trains a character LoRA you own
Built-in disk manager — keep the LoRA, clear the rest
Export the dataset and the trained model
Pricing
Buy once. No subscription, ever.
- Everything, unlimited
- Free updates to this version
- Activate on up to 2 machines
- No subscription, no per-image fees
Requirements
Windows 10 / 11 desktop app.
NVIDIA GPU with 16 GB VRAM.
Works fully offline — no internet needed.
An RTX 4060 Ti / 5060 Ti (or better) recommended.
One license = up to 2 machines.
Buy once. Updates included, no recurring fee.
What's new
Buy once and every update to this version is included.
- 9-step guided pipeline: Hero → Define IP → Drafts → … → Export
- Face-lock editing (Qwen-Image-Edit) keeps one face throughout
- One-click character LoRA training, with live sample shots
- Fully local & uncensored — commercial-friendly models on your GPU
Questions
Is this a subscription?
No — Virtual Human Studio is a one-time purchase. You own this version forever and get its updates. No monthly fee, no per-image charge.
Do I need a GPU?
Yes. It runs on your own NVIDIA GPU with about 16 GB of VRAM (an RTX 4060 Ti / 5060 Ti or better). The heavy steps — generating and training — really want a GPU.
How consistent is the face, really?
That’s the whole point. Step 1 locks a Hero face, every later step is built to keep it, and the LoRA you train reproduces that exact character on command.
Is it uncensored?
Yes. Everything runs locally with no content filters, so what you make is your call — and your responsibility. Nothing is sent to a server to be moderated.
Can I sell what I make?
Yes. The engines (Qwen-Image-Edit, CHROMA) ship under commercial-friendly licenses, and the LoRA, dataset and images you create are yours to use commercially.
Do I own the character and the LoRA?
Completely. The trained LoRA, the dataset and every image live on your disk — copy them, back them up, move them. Nothing is locked to a cloud account.
How long does training take?
A while — it depends on your GPU and how many photos you trained on. The app shows live progress and sample shots so you can watch the face come together.
How many characters can I make?
As many as you like, one IP at a time. Each character gets its own face, dataset and LoRA, kept in its own folder.
Does my data leave my computer?
No. Faces, photos, datasets and models all stay on your machine — Virtual Human Studio runs fully offline.
What about refunds?
If it’s not the right fit, see the refund policy for details.