Train an AI character from a folder of images (or a single picture, or just a sentence) in about 30 minutes, then put them in any image or video you can describe. Built for people who've never trained a model: no nodes, no jargon, and the character is a file you own, not a subscription hostage.
● Every image on this page is a character we trained, made with the tool
Same character ×4The templeMoonlit woodDesert ruinsStorm peak
Not one lucky character
That wasn't a fluke. Here's a completely different person.
Same tool, a face that shares nothing with the first — and a wardrobe and era that don't either. Train an original character, a mascot, a fiction cast, or yourself. Whoever it is holds their identity across every scene, outfit, and century you put them in.
AriesOne character · same tool · ancient to modern
The legionCommandPresent dayWrong century
And not one style, either
Photoreal isn't the only room in the building.
The same trainer makes characters in whatever look your world calls for — crisp modern anime, comic-ink, painterly 2.5D. Pick the style, keep the character.
Any aestheticAnime · comic · 2.5D · realism
Anime · AeliraComic · Freya2.5D · studio
Thirty minutes, three moves
The training other tools make you study, this one just does.
Drop in your images
10–40 pictures of your character: art, renders, or photos. That's the whole requirement. No resizing, no folder rituals. No photos at all? Describe your character in a sentence and OC Studio builds the training set for you.
+
38 images loaded · ready to caption
AI writes the captions
Captioning is the part every guide spends 3,000 words on. Here, an AI studies each image and writes them. You just review. It even knows not to describe the things that make your character them.
MYCHAR in a candlelit temple, ritual markings, dark sash, glowing amber eyes, close three-quarter view…
MYCHAR on a storm-lit mountain peak, tattered cloak, lightning behind, full-body wide shot…
Captioned with AI · editable anytime
One button. ~$3. Thirty minutes.
Training runs in the cloud on open-license models. You watch the progress, then your character appears in your library, ready to generate.
Training your character72%
600 steps · $3.00 · Apache-2.0 base model
Everything after the training
One character. One studio. The whole pipeline.
Once your character exists, you never leave to use them. The five-tool pro stack lives under one roof, and every button shows its price before you press it.
One subscription covers all of it · no seat-fee stacking
GenerateYour character in any scene, any style, any framing.
~3¢ / image
Edit"Remove the sunglasses." Relight, restyle, upscale, cut backgrounds.
from ~0.2¢
AnimateType a scene, get a video clip of your character. No filming.
~25–40¢ / clip
TrainImage and video characters, from photos, a single picture, or a written description.
~$3 · ~30 min
WorkflowsDescribe a ComfyUI graph in plain English. It gets built, then explained node by node.
pay per run
The actual product
Not a mockup. The real studio.
Everything above happens under one roof. Here's what it actually looks like.
Edit by sentence.
"Change the pose to a cross-legged meditation." Inpaint, restyle, relight, upscale, or cut the background — and every version stays in a chain you can step back through.
Any scene, plain words.
Describe it in a sentence and the studio picks the model and settings for you. Your character, any framing, any style.
ComfyUI, without the wiring.
Describe a node graph in plain English. It gets built for you, then explained node by node.
Everything in one place.
Every render, edit, and clip lives in your gallery, tied to the character that made it.
The part nobody else offers
Your character is a file. Not a feature you rent.
Download the model the moment training finishes — it's a standard .safetensors file.
Open-license base models (Apache-2.0). No fine print between you and your own creation.
Works everywhere: keep using it here, or load it into ComfyUI and any other tool, forever.
my-character_v1.safetensors
Type · Character model (LoRA)
License · Yours · Apache-2.0 base
Trained · 28 min · 38 images
Portable · ComfyUI ✓ anywhere ✓
And then they move
Video characters, from text alone.
Train a video version of your character for about $3, type a scene like "she walks through a snowy garden at dusk, slow tracking shot", and get a clip. No filming, no start image required.
One sentence. Their trained model. A moving shot.
This clip is real output — a trained character plus a text prompt, no filming and no start image required.
Generated from text · trained video character
Founder pricing
$8 a month. Locked in for good.
Big suites charge $24–60/mo for generation alone; this is the whole pipeline, training included. Sign up now and lock your rate at $8/month, forever ($10 after launch). Every action shows its cost before you run it, so there are no surprise burns.
Start with 150 free credits to feel it out first — no card required.
