
A trained identity, not a prompt
A dedicated model per trained talent, scored on every frame, with versions you can activate or roll back.
A talent who stays herself · a director you talk to
Nova is a co-director you talk to. Tell her what you are after, out loud if you like. She plans the shoot, runs the studio and answers back. Then she spends your credits as if they were hers: she asks before a render that could go two ways, checks every still against your talent's cards, and fixes seconds, not whole takes.
Coming soonBuilt to publish straight to your channels through KX Bind.Spring launch. Luciana, vertical, one clip for stories.
Suite, rooftop or pool? They look nothing alike, and I'd rather ask now than charge you twice.
Try all three as stills.
Before you spend on video, I checked them against her cards:



Rooftop. Make it a clip.
Shooting it. Those two would have cost you about 900 credits in video.
10:40 AMExample conversation. Checks against her cards are in beta.
Where the money goes
Behind every clip you can't tell is AI are the stills that weren't right and the takes that went wrong on the way to video. Nova's job is to make that pile smaller: narrow the idea before the first render, check before the expensive one, fix seconds instead of takes. Technical failures are never charged. A creative miss is still yours to judge, so she shows you the takes first.
The cheap mistake: a still that was never what you meant.
The dear mistake: a clip paid for on a still that was already wrong.
The take you already paid for: made to count instead of thrown away.
Hands off: say “hey kx” and direct all of it out loud. She answers back.
One talent, ten sets
Luciana was trained once. Every frame here is a separate render, a different set, light, wardrobe and lens, and she is unmistakably herself in all of them.










A character, not a face
She is a person with a bible, not a prompt. Her voice, how she talks, what she would never say and the world she lives in go with her into every still and clip. Everything she makes is seen through that world.

A dedicated model per trained talent, scored on every frame, with versions you can activate or roll back.

Designed from a description or cloned from one you own, with expressive reads, pauses and laughs.

How she talks, what she cares about, what she would never say. Structured sections the whole pipeline reads.

Outfits, products and tattoos you build once and reuse in any scene.

A myth, rules that decide how she looks, the people she knows. Up to six characters share one, and stay coherent across every asset.

Chat and voice calls, as herself, through KX Axis. Her face reacts on the page while the visitor types.
One brief, one campaign
One talent, one conversation, first rough ask to final set. Nothing here needed a camera, a set or a second day.
“Spring launch for the skincare line. One hero image, three vertical clips, a voice-over. Due Friday.”
Suite, rooftop or pool? They look nothing alike, so she asks. Product in hand in every frame, cyan wardrobe, and the price named before anything renders.
Each scored against her identity and compared with her cards. Two flagged, one kept, before a clip is paid for.



The rooftop still becomes a 9:16 clip. Nova watches it: at second 4 the product changes hands. She rewrites the beat and re-renders those seconds only.

Hero still, clips and the voice track, versioned in the project and downloadable today.




Why not a creative suite
Good suites plan from a chat, pick the models and keep a library. None of them know your talent's cards, remember what a wording did last time, or take direction out loud.
| Aspect | A conventional AI creative suite | KX Visuals |
|---|---|---|
| Starts with | A prompt, or a chat that picks the tools | A talent who already exists, and a conversation about what she should do |
| Before production | Executes the request | Asks when the brief could go two ways, and names the price |
| A wrong take | Generate again | Caught against her cards before the render, or re-rendered for the seconds that failed |
| Identity | Saved characters and reference consistency | A trained model, scored 0–100 after every generation |
| The talent | A saved look | A look, a voice, a lexicon and a world she is seen through |
| Memory | A shared library | The brief, the decisions and what each wording did last time |
| Directing it | Type | Say “hey kx” |
The engines
Nova routes each job to the best current model and keeps the identity, the brief and the continuity constant across them. Models change. Your talent does not.
Image-to-video and the default for looping clips.
Motion-heavy clips and extended takes.
A second opinion on motion when the first take drifts.
First call for plain-language edits.
Multi-reference edits that keep the face, product and room.
Sets, backgrounds and the trained identity itself.
Scripted speech on a clip, compared side by side.
Designed or cloned voices, expressive reads, sound effects.
Nova's reasoning, and the judge that scores every frame.
Engines are swapped as better ones ship. The identity, the brief and your project memory stay.
For other software
Everything Nova can do, your own software can ask for: the same tools over MCP and A2A, and a headless pipeline that composites brand creatives for other KX products.
The studio’s tools, callable by your own agents with API authentication. Cast, shoot, check and fetch the result without opening the console.
A headless path for backdrops and composited creatives, already used by other KX products.
Public and member delivery tiers for the stills and clips a talent produces, so her content lives inside your app.
Pricing
Every plan includes Nova, the whole studio, a cost estimate before every render and no charge for technical failures. Credits go on approved shoots, and Nova is built to waste fewer of them.
Free today. No card, no meter.
$0
400 credits a month at launch≈ 13 stills, or 1–4 clips
Build one recognizable talent and keep creating.
$19/mo
2,000 credits a month≈ 65 stills, or 4–19 clips
Continuous content for a creator or a brand.
$45/mo
5,500 credits a month≈ 180 stills, or 12–52 clips
Several campaigns and a small team.
$119/mo
16,000 credits a month≈ 530 stills, or 35–150 clips
Talent and campaigns across clients.
$249/mo
38,000 credits a month≈ 1,250 stills, or 84–360 clips
Approximate. A still is about 30 credits; a five-second clip is 105 to 450 depending on the engine Nova picks, and she shows the exact figure before every render. The numbers firm up as usage is measured.
Launch pricing. Nothing is metered or charged until self-serve billing goes live; until then every account is free. Coming soon Self-serve plans and billing are on the roadmap.
Questions
A studio you talk to. You train one photoreal talent for your brand, and Nova, the co-director inside it, plans and produces on-brand photos, short video and voice with that same face in every frame. No photoshoot, and fewer wasted renders.
Those give you every model and a prompt box; you do the directing and pay for each guess. KX Visuals starts from a trained identity, keeps a brief and a memory of your brand, and puts Nova in charge of routing, running and checking the work. You decide what to make. She works out how.
Three things, at three moments. Before the first render she asks when a brief could go two ways and names the price. Before the expensive one she rewrites wording the engine misreads, compares props and wardrobe with their cards (in beta) and scores every frame against the identity. After a take she names what read false and when, writes the fix, and re-renders only those seconds.
Yes. Say “hey kx” in the console and direct the shoot out loud; Nova answers back in her own voice. Typing still works, and her questions come with chips you can tap.
Soon. KX Axis lets her answer chat and voice calls as herself, in her own voice and lexicon, with her face reacting on the page. Today her reaction loops are built and packed for chat; the channel is on the roadmap.
That is the product. Each talent has a dedicated trained model, every new photo is scored 0–100 against the reference identity, and frames that drift are flagged and remade. Permanent markers such as tattoos or scars are enforced on every render.
Yes. Looping vertical clips with sound are live. Designed or cloned voices with expressive reads are live. Lip-synced talking video is in beta and may need a hand from us.
Several, and they change. Today that includes Seedance, Kling and Wan for video, GPT Image, Nano Banana and FLUX for stills, Sync for lip sync, ElevenLabs for voice and Claude for Nova's reasoning. Nova routes each job; you never have to pick.
Sign-up is open and the Explore plan is free today. Launch plans start at $19 a month with credits included; billing goes live with self-serve plans, and nobody is charged until it does.
Yes. Every asset is produced from an identity you own and rules you set. The full terms are being finalized with counsel; the working draft is linked in the footer.
Four talents, no photoshoots
See what people make →



Sign up free and direct your first shoot with Nova today. Custom-trained talent and agency workspaces are set up by hand.
Start free