Kaleido keeps templates, AI generation, batch jobs, and retained assets inside one workspace. Images scale out concurrently; video workflows open up gradually under a serialized QA model, with every step visible.
Generate a summer commute ad set for this travel coffee cup, tailored for Meta and TikTok.
Two production rhythms
Instead of forcing every task into one prompt box, Kaleido gives different cost, latency, and QA models the workflow each one actually needs.
Pick a template, swap the product, tune the copy, and submit multiple variants at once. Results stay easy to compare, retry, and export.
The plan is either template-led production or Agent-assisted shot planning, executed one task at a time. When video capabilities are connected, queueing, progress, and failures stay just as explicit.
From input to reusable assets
Inputs do not disappear after one chat, and outputs do not get lost in a downloads folder. Every step returns to the workspace and asset library.
Upload product images, reuse saved subject assets, and add missing details manually whenever a product link cannot provide them yet.
Templates hold the controllable structure while the Agent organizes selling points, compiles prompts, and prepares batch work.
Track the real status, retry only the failed item, and keep every result in the asset library for the next round.
Built for daily output
Each capability exists to help teams start faster, generate more reliably, and reuse assets at lower cost.
Apply one structured template to the same product image and generate up to 10 variants in a single concurrent batch.
Subject isolation happens before scene generation so the model does not casually redraw the product itself.
Text is laid out with Canvas / HTML instead of asking the model to render lettering, making edits and language swaps easier.
Re-layout and export for common ad frames such as 1:1, 9:16, and 4:5.
Products, human references, and generated outputs stay stored as assets ready for the next creative cycle.
The Agent helps organize selling points and prepare prompts, while autonomous execution opens up gradually as integrations land.
Honest by design
Kaleido treats task state as part of the product, not as an implementation detail hidden in backend logs.
Questions
Today, Kaleido is aimed first at cross-border media buying teams, optimizers, and ecommerce operators who need a steady stream of ad creatives. The goal is not a general AI chat tool; it is a production flow that connects templates, batching, queues, and asset reuse.
Images are cheaper and faster, so they fit concurrent generation with a cap of 10 per batch. Videos cost more, take longer, and need tighter QA, so they run strictly one by one with explicit queued, running, succeeded, and failed states.
No. Kaleido treats copy as a separate layout layer so teams can keep text sharp, edit it manually, swap languages, and reflow across different aspect ratios.
Not yet. The current product is a pure creative generator focused on input, generation, review, export, and asset retention. Media buying and performance feedback belong to a later stage.
Move from idea to output
Turn one set of product inputs into reusable creative output. Start in Studio with the image workflow or collaborate with the Agent.
Enter Kaleido Studio