Docs AI mode
AI mode
With ai=1, a language model art directs the name before the engine draws it.
B Klug, updated October 25, 2026
The standard engine picks fonts, colours and symbols by chance, guided by the tone and industry it reads from the name. AI mode adds a step before that. Add ai=1 to a logo request and Claude Haiku 5.5 reads the name and writes a short brief: what the business probably is, and six distinct directions for it. Each direction fixes the palette, the background, the typefaces, the case, the layout and a symbol family. The engine then draws the logo the direction describes, using the same catalogue and the same renderer as before.
https://api.logolava.com/v2/logo.png?text=Lighthouse+Yoga&ai=1&seed=387438933&w=540&h=200
To see the six directions for a name before writing any code, try the AI logo generator.
Six directions, picked by seed
The brief holds six directions, and the seed picks one of them. Seeds that differ by one step land on different directions, so a handful of seeds shows the whole range for a name. Because the brief is stored the first time it is made, every later request for that name reuses it. The same name, options and seed give the same logo again, with no new model call.
You can read the brief itself at /v2/brief?text=Lighthouse+Yoga. The response includes the model, a reading of the name, the detected industry and tones, a rationale, and the six directions with their concepts, colours, fonts, layouts and symbol families.
Options
| Parameter | What it does |
|---|---|
ai | 1 turns AI mode on. Without it, nothing here applies. |
aiset | A number from 1 to 20. Each set is a different brief of six directions for the same name. The default is 1. |
hint | A short free text about the business or the mood you want. It is passed to the model as data and never as instructions. |
tagline | Set to ai to have the model write a short tagline under the name. |
aimark | 1 lets the model draw a custom mark from simple shapes instead of choosing a catalogue symbol. This is experimental. The shapes are checked, and the catalogue symbol is used if they fail. |
Reading the status
Every AI request answers with an X-LogoLava-AI header. fresh means a new brief was just made. hit means a stored brief was used. off means AI mode is switched off. A value starting with fallback: means the brief could not be made, so the standard engine drew the logo, and the reason follows the colon. The same message appears in X-LogoLava-Warning. You always get a logo; it just may not be an AI directed one.
Limits
A new brief costs a model call, so briefs are limited more tightly than ordinary logos. Without a key you get 20 new briefs a day and one every 30 seconds for each address. With a free API key you get 300 a day, with a gap of 2 seconds. A logo request that arrives inside the wait gets a standard logo, with a fallback status in the header. The JSON forms add an aiRetryAfter value in seconds, and /v2/brief answers 429 with Retry-After. A brief that already exists costs nothing and never waits, so a page of logos with fixed seeds is safe to show to any number of visitors. The full table is on the limits page.
A fresh brief needs a model call, so it takes longer than a stored one. In an app, ask for the brief when the person enters a name, and show the standard logos in the meantime.
Next: Catalogues