JuoniImport · ChatGPT
IMPORT · CHATGPT

Turn a design you made in ChatGPT into a presentation.

Designed the deck in ChatGPT? Ask it for the file and drop that on New. The same reader that takes a Claude Design takes it: your text stays text, your shapes stay shapes, the colors arrive as a style you can keep, and you see exactly what came across before anything is created.

01

The three steps.

The whole import is three moves. The first is the one that is different from Claude Design: ChatGPT has no export menu, so you tell it what file you want, and the prompt from New tells it what to put in that file.

  1. 01
    Paste the prompt from New into ChatGPTOn New, under Still designing it, switch to ChatGPT and copy the prompt. It tells ChatGPT the slide canvas, the grid, your fonts and every setting the editor can read. Design the deck there.
    Design it with the import in mind
  2. 02
    Ask for the deck as one .html fileOne file, with the styles inline, is what the reader wants. If the design has pictures, ask for a .zip with the images beside the .html, so they arrive as files you can swap.
    Or a .zip with its images
  3. 03
    Drop it on NewYou get a report before anything is created: what mapped, what did not, and what to do about it — per slide. Then open the editor.
    Then it is a deck like any other
FIG 1

The well on New. The same well takes a Claude Design and a PowerPoint deck; the report names which arrived.

02

Why one file, and why the prompt.

The importer is a reader of HTML. It measures the page as a browser would draw it and lands each kind of thing as the kind of thing the editor already has. That is why the file to ask for is one .html with its styles inline, and not a picture, a PDF or a slide file: the structure is what carries across, and a picture has none.

The prompt on New is how ChatGPT learns what that page should be. It names the slide canvas, the deck’s twelve-column grid and its margins, your kit’s fonts, one section per slide with a label and speaker notes, real text and never a picture of type, figures in a real table, and nothing that runs. It also names every setting the editor can read and what each snaps to, so a gradient or a shadow ChatGPT sets arrives as the same setting here. Paste it before you start, and the design comes across as native slides rather than a drawing of them.

What to check before you ask for the file

  1. Real text, not pictures of text. Words in an image stay an image.
  2. One design per file. Two designs in one page import as one deck.
  3. Pictures as image files beside the .html, one per picture, in a .zip.
  4. Figures in a real table, not a chart drawn on a canvas.
03

What comes across.

Each kind of thing in the design lands as the kind of thing the editor already has. The table is the same one the Claude Design guide shows, because it is the same reader.

In the designIn the deckYou can
TextHeadings, paragraphs, listsEditable text, on the deck’s type stylesRetype it, restyle it, ask the agent to tighten it
Boxes, borders, rulesDivs with fills and edgesNative shapes and linesMove and recolor them like any block
ImagesThe files in the zipPictures carrying the fileSwap them, or reuse them in any deck
ColorsThe palette ChatGPT choseA style for this deckKeep it, or switch the deck to your brand kit
Gradients, corners, shadowsAs set in the CSSSnapped to the style’s steps, still editableAdjust them in the panel
TablesA real <table> of figuresAn editable table, its header band and column widths keptTurn it into a native chart in one click
Labels, speaker notesA label and a note per sectionSlide titles and the notes panePresent from them
Scripts, animation, canvasAnything that runsA flagged box at its exact sizeReplace it with the end state you wanted
TABLE 1

The mapping the importer applies. Nothing on the left is rasterized; everything in the middle stays editable.

04

What to look at afterwards.

A design is free geometry; a deck is a grid. Blocks that landed off a column or in the margin are flagged on the slide, and Snap to grid puts them on it in one move. A block that reaches the edge of the slide is taken to be bleeding on purpose and left alone. If ChatGPT used a face that isn’t one of the deck faces, the style names the one it picked, so you can see what changed before anyone else does.

The same well takes a Claude Design and a PowerPoint deck.

OFF-GRID BLOCKS ARE FLAGGEDSNAP TO GRID IS ONE MOVEBLEEDS ARE LEFT ALONENOTHING IS METERED
QUESTIONS

Short answers, one each.

Anything longer is a section above.

01

Does ChatGPT need a plugin or a connection to Juoni?

No. The design leaves ChatGPT as a file you download — one .html, or a .zip with its images — and you drop that file on New. Nothing is connected and nothing is sent from ChatGPT to Juoni.

02

Why one .html file and not a PowerPoint or a PDF?

The reader reads the structure of an HTML page, so text stays text, shapes stay shapes and the colors become a theme. A PowerPoint made by ChatGPT can be imported through the same well, but a PDF or a screenshot would arrive as a picture you can only crop.

03

Is importing from ChatGPT different from importing from Claude Design?

Only in how the file gets out. Claude Design exports Project HTML from its Share menu; from ChatGPT you ask for the file. The same reader reads both, the prompt is the same prompt, and the mapping is the same table.

04

What if the design has pictures?

Ask ChatGPT for a .zip with the images beside the .html, and reference each as an image file rather than a CSS background. Pictures in the zip arrive as assets you can swap; a picture the file does not carry comes in as an empty, flagged box.

05

Does importing a ChatGPT design use credits?

No. Importing and editing are never metered, on any plan. Credits are spent only by agent turns and generated pictures, and the editor quotes each one before it runs.

Start with a sentence. Keep the deck.

Whatever the agent builds is yours to take apart — every slide, every box — in an editor that was there first.