Recipes
Pick a recipe, answer a few questions, press once. The model, the settings and the prompt somebody spent a week getting right are already decided, so all you bring is what only you know: your product, and how you want it shot.
What it already knows
{{creative_prompt}}your answer{{slot_instructions}}your picture{{choice:angle}}your pick{{choice:background}}your pickThe prompt's wording stays with the recipe. Your answers fill the gaps.
What it asks you
Listing ShotReady
Create·3
Four questions, then it runs, for 3 credits: the same as typing the prompt yourself.
Three kinds
In the catalog all three look the same: a card, a few questions, a price. What differs is what happens when you press it.
Drops your answers into one prompt and makes one thing.
The recipe above is one, and so is most of the catalog. The model, the settings and the wording are decided; your answers fill the gaps the author left.
Best forA look you want again and again, like the same clean shot for every product in your shop.
Has an AI look at your photos and write the prompt for them, then makes one thing.
The same questions as a Fill-in recipe, but the author wrote directions instead of a prompt. An AI reads them, looks at the photos you brought, and writes a prompt that fits those photos. Writing it costs you nothing.
Best forPhotos that differ a lot, where one fixed prompt would suit some and miss the rest.
Runs several steps in a row, and can make many things from one press.
Steps that feed each other: write a script, voice it, draw a picture for every scene. One step can make one thing for every item in a list. The rest of this page is about this kind.
Best forA batch from one brief, or a finished piece — a storyboard, an edit — that takes several steps to make.
Workflows
A workflow is drawn as a board. The questions it asks sit in one column and the steps it runs in the other, in order, and a wire from one card to another means “this step uses that”. There are no if-this-then-that branches: the steps simply run top to bottom.
This board was built in Kuro. Paste a narration script, pick a voice and your characters, press once, and a storyboard arrives with the voice-over already laid under it.
Sound The narrator reads your script aloud, in the voice you picked.
Transcribe The recording is timed line by line, so each scene can last exactly as long as its words.
Write An AI cuts the transcript into scenes and writes, for each one, what the picture shows and how the camera moves.
Image One still per scene, with your characters' faces. This step repeats: twelve scenes, twelve stills, one press.
Storyboard It all lands on a board in the Studio with the voice-over underneath. Fix what's wrong, then turn the scenes you like into clips, one at a time.
Point a step at a list and it runs once per item. Six prompts become six images from one press, and each one lands in your library like anything else you make.
Each result is saved the moment it's made. If a run stops halfway, it picks up where it left off: you lose time, never credits for the same picture twice.
Six kinds of question and ten kinds of step. Any board you build is made of these.
Up to 8 questions, answered once before the run starts.
The eight things it can ask for
An image, a clip or a sound — up to 8 — from your uploads or anything you've already made.
Free text, used by any step that asks for it. The author decides how long it can be.
One exact line per carousel page, kept separate from the generated pictures.
Choose a 9:16 or 4:5 page for a still-image carousel.
A number within limits the author sets. Usually it is “how many?”, which is what a step repeats over.
A voice and the engine behind it, picked by you when you run it rather than fixed by the author.
Characters, places or products saved in your account. Tag one and its face and description come along.
A short list of models the author picked, up to 12, so you can trade quality for price without touching the board.
Steps run in order, and each one names exactly what it uses: one of your answers, something an earlier step made, or the item it is repeating over. A board that asks for something nothing makes can't be saved.
The ten things it can do
Roughly in the order you'd use them: read what came in, write, make pictures, clips and sound, then finish.
Finds the cuts, key frames and pacing of a video you bring, then groups its shots into beats. The timings are measured from the video, not guessed.
An AI writes text: one line, a list, or a set of filled-in fields. It can look at the run's pictures. This is how one brief becomes six prompts.
Makes an image exactly as the composer would: same credits, same library. Point it at a list and it makes one per item.
The same, for video: one clip per item, on the model the board names or the one you picked.
A voice-over or a music track, from a script and a voice.
A finishing tool: remove a background, upscale, join clips, burn in subtitles.
Turns speech into timed text. Transcribing the same file twice costs nothing: the first answer is kept.
The run stops and waits for you. It can show you up to 4 of the things made so far, and tell you what finishing will cost before you say yes.
Edits up to 6 tracks together on a 9:16, 1:1 or 16:9 canvas: clips in sequence, a soundtrack at the volume you set, subtitles in the style you pick.
Lays a video out as scenes on a board, each with a title, a still and a camera move. It costs nothing by itself: you make each scene's clip in the Studio when you're ready.
Wire a later card into an earlier one and the board moves it behind what it uses. You can't draw a board that couldn't run.
Limits and runs
The limits depend on whose workflow it is. Your own get the high ones. Somebody else's published workflow gets the lower ones, because it is spending your credits.
| What a workflow may do, and whose credits it spends. | A workflow you wrote | A workflow somebody published |
|---|---|---|
| Steps | 20 | 6 |
| Steps that write | 5 | 3 |
| Things one step may make | 30 | 10 |
| Runs at once | 3 | 3 |
“Steps that write” is the one limit that isn't about your bill: writing runs on our AI and costs you nothing, so it is capped on its own.
Pressing a workflow starts a run you can follow. The board redraws itself as it goes, and each card shows what it made.
A saved setup for making something in Kuro. The model, the settings and the prompt are decided, and it asks only for what it can't know, like your product photo or a line about it. You answer, press once, and the result lands in your library.
No. A Fill-in or Reprompt recipe costs the credits of the one thing it makes, the same as asking for it yourself, and having an AI write the prompt is free. A workflow costs the credits of what its steps make, and its writing steps are free.
Yes. In the app, New recipe asks which of the three kinds you want and opens its builder. You can also turn any image or clip you've made into a recipe that makes more like it, or copy one from the catalog and change it. Every recipe you make is private until you publish it.
Yes. Claim a handle, give the recipe a cover and publish it: it gets its own page in the public catalog with your name on it, a link you can put anywhere. Every recipe is checked before it goes live.
The catalog
Every public recipe, in one place. Browsing is free and you only spend credits when you run one. Or build your own, and publish it under your name.