
Studio generation
A full AI-native studio for AI UGC, ads, clipping and avatar creation.
NEW Images Agentic video API
Private AI in your browser. Use Seedance, Wan, DeepSeek, GLM 5.3 fully uncensored and abliterated.
Arcana Auto
Here it is. I kept the palette cool and let the beam cross the moon.
Arcana Auto
It waits until calls stop for ms milliseconds, then runs fn once:
export function debounce<A extends unknown[]>( fn: (...a: A) => void, ms = 200, ) { let t: ReturnType<typeof setTimeout>; return (...a: A) => { clearTimeout(t); t = setTimeout(() => fn(...a), ms); }; }
Arcana Auto
Yes. Wrap the handler once and pass the result to addEventListener, so a burst of resize events runs your code a single time.
Text chat with folders and search. History is encrypted in your browser.
One web app for chat, images and video, with an API on the side. Your chat history stays encrypted on your device.

A full AI-native studio for AI UGC, ads, clipping and avatar creation.
Run them from one chat - agents, images and video included.
No server-side chat history and no account sync. Clear it any time from Settings.
Every request goes through the Arcana relay, so the model provider sees Arcana, not you. The relay keeps no copy of your chats, and your history stays encrypted on your device.

And add him in whichever scene you like.
| Provider | 480p | 720p | 1080p |
|---|---|---|---|
| Arcana | $0.0988 | $0.2196 | $0.5402 |
fal | $0.2205 | $0.4730 | $1.1640 |
Higgsfield | $0.2057 | $0.4622 | $1.1373 |
Sume | $0.1028 | $0.2312 | $0.5686 |
| Provider | 480p | 720p | 1080p |
|---|---|---|---|
| Arcana | $0.0500 | $0.1000 | $0.2000 |
fal | not shown | not shown | not shown |
Higgsfield | not shown | not shown | not shown |
Sume | not shown | not shown | not shown |
List prices in USD per generated second, no reference video, checked : Arcana Studio, fal, Higgsfield, Sume.
Short answers about privacy, pricing, models and the API.
Ask us something elseOn your device. Conversations, folders and drafts are encrypted and saved in your browser, and are not synced to an Arcana server.
The message you send and the context the model needs to answer it. Arcana relays that request to the model and passes the reply back. It does not keep a copy of the conversation.
There is no monthly plan. You top up Tokens and each request spends from that balance, priced per model. The Tokens page shows every call and what it cost.
Text, image and video models from one catalog, including Seedance and Wan for video. You can switch the model for each chat, even halfway through a conversation.
Yes. Ask for what you want in plain words. The chat agent writes the prompt, starts the job and drops the result into the conversation when it is ready.
Longer or batch work: several takes of one idea, image-to-video, reference clips. Studio keeps a queue and a history of your generations.
Create a key on the API page and point any OpenAI-compatible SDK at Arcana's base URL. Chat, image and video calls draw from the same Tokens balance.
Your local chat history is removed with it, because that is the only place it lives. Your Tokens balance and API keys are not affected.
Chat, images and video in one place. Your history stays in this browser, and you pay only for what you use.