Use cases › Document import
The document already says it. Now it has to present.
Upload a PDF, a Word file, an existing deck, a transcript or a Markdown file. What comes back is a structured presentation, not a page of your report shrunk to fit a slide.
There is a specific kind of Tuesday where the work is already finished and the presentation is not. The analysis is written. The report has been reviewed. Somebody now needs twelve slides of it by Thursday, and the honest version of that task is reformatting rather than thinking.
The application reads PDF, Word, PowerPoint, plain text and Markdown. It parses the document into sections, decides what each section is arguing, and lays that out as slides with a structure rather than as paragraphs in a smaller font.
The failure mode everyone has seen is a slide with four hundred words on it, produced by a tool that treated conversion as a copy operation. Generating from the structure of the document instead of the text of it is what avoids that, and it is why the output is shorter than the input.
Grounded mode, which matters most here
When you attach a document you can require that the deck invent nothing. No figure, percentage, currency amount, date or quotation reaches a slide unless it appears in your source, and whatever gets dropped is reported back to you. If a claim in the document has no number behind it, the deck does not acquire one on the way out.
Grounded output follows the document's shape, at roughly one slide per section, so the length comes from the source rather than from a number you picked.
What a good one contains
Six slides that earn their place, and what belongs on each.
What it is and what it reads
PDF, DOCX, PPTX, TXT and MD. A research report, a memo, a transcript, a spec, a set of meeting notes.
A clean heading structure in the source produces a better deck. Headings are the outline.
Sections become slides The document is parsed into sections and each one is assessed for what it is arguing, not just what it says.
This is why the deck is shorter than the document. Most paragraphs support one point.
Numbers stay attached to their source With grounded mode on, every figure is traceable and anything unverifiable is stripped and reported.
The removal report is the useful part. It shows which claims were never evidenced.
A layout per slide, not one template repeated Comparisons get a comparison layout, sequences get a timeline, a single number gets a slide to itself.
Bulleted lists on every slide is the tell of a converter rather than a generator.
Set the audience The same document aimed at the board and at a customer produces two different decks.
Executive team, Board of directors, Investors, Sales team or Customers.
Export to a real file PowerPoint with editable text boxes, PDF, or one self-contained HTML file that presents offline.
You can also push it on to Google Slides or Canva.
How it goes
PDF, DOCX, PPTX, TXT or MD
The document you already wrote, not a summary of it. More context produces a better structure.
Require it to stay inside the source
No invented figures, and a report of anything removed for being untraceable.
Choose the audience and the length band
The register changes with the audience. The length band sets how much gets cut.
.pptx, PDF, HTML or a share link
Editable text boxes in the PowerPoint file, so the next edit does not come back to us.
Before and after
Same content. One was written at 1am in a text editor.
3.2 Market conditions
- The mid-market segment has demonstrated consistent expansion across the trailing twelve month period, with particular concentration in the healthcare and logistics verticals, where procurement cycles have shortened considerably relative to the prior year, although this has been partially offset by...
A paragraph from the report, pasted onto a slide and shrunk until it fits. This is what document conversion usually means.
Mid-market grew fastest where buying got quicker
Healthcare and logistics, trailing twelve months
The same section, read for what it argues. The detail moves to the appendix, where a question can still reach it.
Questions
Which file types can it read?
PDF, Word (.docx), PowerPoint (.pptx), plain text and Markdown. A research report, a memo, a meeting transcript, a specification or a set of notes all work as input.
Will it just paste my paragraphs onto slides?
No, and avoiding that is the main design decision. The document is parsed into sections and each section is assessed for what it argues, then laid out as a slide with a structure. That is why the deck is considerably shorter than the document it came from.
How many slides will I get?
With grounded mode on, roughly one per section of your document, so the length follows the source rather than a number you chose. Generating without a source document lets you ask for a specific count instead.
Can I trust the numbers it pulls out?
With grounded mode on, every figure on the finished deck traces back to a line in the file you uploaded, and anything that could not be traced is removed and listed for you. Without it the model is free to write, which is useful for a first draft and wrong for a board pack.
Can I turn an existing PowerPoint into a better one?
You can use a .pptx as a source document and generate a new deck from its content. What is not built yet is reading a brand out of that file and applying it, so the new deck is styled with one of the five built-in themes.