EditableImage

Guides · · Updated · By River · 9 min read

How to convert a screenshot to an editable PowerPoint: 3 methods compared

Rebuild it by hand, OCR the words and lay them out again, or let a converter reconstruct the slide. One CC-licensed slide through all three, with the files to check.

The test slide screenshot beside the same slide rendered from an exported PPTX, under the heading Screenshot to editable PowerPoint, 3 methods, one slide

A screenshot of a slide is one picture. To edit the words on it you have three options: rebuild the slide by hand in PowerPoint, run the picture through OCR and lay the words out again, or hand it to a converter that reconstructs text boxes, pictures and a background in their measured positions. The first costs the most time and gives you full control. The second saves the typing but none of the layout. The third gives you a slide in under a minute that still has to be checked, and this guide shows exactly what needs checking.

To compare them on the same ground we took one slide from a publicly licensed deck, rendered it at page width the way a screenshot of the PDF would look, and put the identical file through all three — including six online converters on their free tiers, two of which produced a file without an account. One of those two is ours, which is the reason to show the other.

The test slide

The source is slide 10, Considerations, from Ismael Olea's talk Wikibase: State of the Cite at WikiCite 2025, published on Wikimedia Commons under CC BY 4.0. We rendered page 10 of the PDF at page width to a 1280×720 PNG, which is what a screenshot of that page looks like at 100% on a laptop.

The test slide: a red title Considerations, two levels of bullets, a four-by-four table of ticks and crosses comparing Wikidata, WB.cloud and WB Suite, and the WikiCite Bern '25 logo
The input, 1280×720. Download the PNG and run it through anything you like.

It is an ordinary working slide, which is the point: a red title, two bullet levels, a four-column table whose cells are ticks and crosses, a logo, and one glyph the deck's own font could not draw — the arrow after LOD shows as a box. The PDF carries the real text, so every method can be scored against it: 30 ordinary words, plus the ticks, crosses and bullet marks.

Method 1: rebuild it by hand in PowerPoint

  1. Insert the screenshot on a blank 16:9 slide and send it to the back, or keep it on the next slide to glance at.
  2. Add a text box for the title, one for the bullets, and a 4×4 table from Insert → Table. Type everything in.
  3. Match size, colour and spacing by eye; set the tick and cross cells in a font that has those glyphs.
  4. Insert the logo as a picture — crop it out of the screenshot or fetch the original.
  5. Delete the reference screenshot.

What you keep: everything, because you chose it — real bullets, a real table that can take a new row, fonts you picked. What it costs: every word retyped and every box placed by hand, and the spacing is yours to get wrong. Nothing on this slide is recovered; it is remade. For one slide that is a fair trade. For a forty-slide deck it is a day.

Method 2: OCR the words, then lay them out again

OCR reads the words out of the picture so you do not retype them. We ran the free, open-source Tesseract 5.3 on the screenshot with default settings; it took two seconds.

It read 28 of the 30 words correctly, and the two it missed were run together as allinfo. What it lost is everything that is not a word: the bullet marks came back as e and °o, the ticks and crosses as Vv and x, the table as four lines of words separated by spaces, and the title's colour and the bullet levels did not exist for it at all. The output is a text file.

That is the honest shape of OCR for slides: it saves the typing and keeps none of the layout. Pasting the text into PowerPoint leaves you at step 2 of the hand rebuild, with the table still to draw.

Method 3: let a converter reconstruct the slide

A converter reads the layout as well as the words: it measures where each line sits, how big, bold and what colour it is, places a text box there, cuts pictures out as pictures, and paints a background for the gaps. On 31 October 2026 we tried six on their free tiers, default settings, same PNG:

  • EditableImage — Screenshot to PPT, 4 credits. We built it. The run below went through the same code and models that serve the site, from a developer build, so that we could time it without the live credit meter in the way.
  • 2pptx — 2pptx.com, Standard tier, labelled "free forever", no account asked for. Paid tiers at $0.08 and $0.40 a page promise cleaner backgrounds and movable elements; we did not test them.
  • SlidesPilot, SlideRebuild, Codia NoteSlide and LayerPSD all accepted the upload, or advertised a free run, and then stopped at a sign-in wall: Google, Apple or email for SlidesPilot, email for SlideRebuild, Google only for Codia and LayerPSD. Free, with an account, is a fair offer; it is not the same as free, so they are out of this round.

What the two files contain

EditableImage 2pptx (Standard)
Time to a downloadable PPTX 25 s 30 s
Words right, of 30 19 as text, every one correct; the 11 inside the table stay in the table picture 30, as 22 loose boxes
Text boxes 5 22
Pictures 3: the background, the table and the logo 1: the background
The table One picture, in place, exact Every cell its own text box; no table object
Ticks and crosses Exact, inside the table picture Three ✔ left in the background picture; the rest became V, 1 and X as text
Bullet marks In place The letter O in its own text box
The logo A picture, cut from the original Read as text: ikite and a stray Chinese character
The missing-glyph box after LOD Kept as a box, as on the slide Dropped
Title red, body grey Kept (measured) Kept (measured)
Font sizes 19–33 pt, measured per line 14–35 pt, measured per line
Watermark None None in this run

EditableImage's column is the re-run of 1 November, after two fixes the 31 October run prompted; 2pptx's is its 31 October run. Both of EditableImage's exports are linked below, so the before and the after can be opened side by side.

The EditableImage export rendered: the title and bullets as text in matching sizes and colours, the table as one exact picture, and the WikiCite Bern '25 logo as a picture in its corner
EditableImage's PPTX as LibreOffice renders it: five text boxes over the rebuilt background, the table and the logo as pictures. Download the file (72 KB); the 31 October export (49 KB) and its render are kept for comparison.
The 2pptx export rendered: every word as text including each table cell, bullet marks as the letter O, two ticks replaced by V and 1, and the logo replaced by the word ikite in large type
2pptx's PPTX as LibreOffice renders it. Download the file (37 KB).

What to look at in the files

EditableImage gives back the slide as it was, with the words that can be retyped as text and everything else as pictures in their places. The title and the four bullet lines are text boxes at the measured sizes and colours, every word as the author spelt it — hypothesys included; the bullet marks stay where they were; the table is one exact picture, ticks and crosses untouched; the logo is a picture cut from the original; even the box the deck's own font drew for a missing glyph is kept, because it is what the slide shows. Nothing is invented and nothing is re-set in a face it was not in. The limit is the table: kept as a picture, its words are exact but not editable, so a reader who must change crow sourcing retypes inside the table.

Two of those results are new. On 31 October the export had read the logo as the characters the model saw on it and had "corrected" hypothesys to hypotheses; both were fixed on 1 November by letting a second reading already in the pipeline overrule the first — the character-level reader's spelling where it saw the type large and is sure of the word, and the segmenter's logo outline where it covers the parser's text — and the slide was re-run. The 31 October export is linked above for anyone who wants to check the difference.

2pptx read every word, including the table's, and that is a real advantage when the words are what you need to change. The cost is structure: the table is 13 loose text boxes aligned by position, three of its ticks are left behind in the background picture where no text box can reach them, the bullet marks are letters that will not behave as bullets, and two of the ticks are now a V and a 1 in a bold italic face. Its logo came back as text in large type. Nothing in its free export was watermarked.

Where both stop

Neither produces a table object: one keeps the table as a picture, the other scatters its cells as boxes. Neither keeps the deck's typeface — both set the text in a substitute and measure its size, which is all a screenshot allows. Both deserve the author's eye on every symbol before the slide is shown, because a tick that becomes a 1 is wrong in a way that reads as right.

Which method, when

  • One slide, you know the fonts, it has to be exact: rebuild by hand.
  • Many text-heavy slides and time to lay them out: OCR, then rebuild, with the screenshot beside you.
  • A deck that survives only as screenshots and the next edit is a title, a label or a figure: a converter, then the checks below. From this round, EditableImage when the slide has to come back as it was — words spelt as written, bullets, ticks, logo and figures in place, nothing invented — and you can retype inside a table yourself; 2pptx when the words inside a table are the one thing you must change and loose text boxes will do.

Capture the screenshot so it can be read

One slide per capture, at 100% zoom or in full-screen presenter view, saved as the PNG your screenshot tool makes, with the window chrome and the thumbnail strip cropped out. From a PDF, capture at page width. A capture from a chat app has been compressed once already; zoom in on the smallest label first. The screenshot-to-PPT page goes through each source — chat, PDF, recording, projector photo — and what each one costs.

Check the export before you present it

Whatever produced the file, open it in PowerPoint and do five things, in this order: read every number, label and symbol against the screenshot; look for headings that wrap a word early; drag one text box and one picture to confirm they are separate; treat any chart or table that came back as a picture as a picture; view the slide at 100%. The five tests guide says what each one catches.

Questions

Can PowerPoint itself turn a screenshot into editable text?
Not as a whole slide. PowerPoint inserts the picture as one object, and while recent versions can lift text out of a picture into a text box, every box, bullet level and table cell still has to be placed by hand. That is the first method below.
Which method keeps the original font?
None of them, because a screenshot holds letter shapes and not a font file. Rebuilding by hand lets you choose the right font if you know it. OCR and converters set the text in a substitute; the converters at least measure size, weight and colour from the pixels, and both we tested kept the red title red and the body grey.
Does a table in the screenshot come back as a PowerPoint table?
Not in either converter we ran. EditableImage returned the whole table as one picture, so its words cannot be edited; 2pptx returned every cell as its own text box, so the words can be edited but there is no table object to add a row to. Rebuilding by hand is the only route to a real table.
Is a free converter good enough for a one-off slide?
For the words, yes — 2pptx read all 30 of them on its free tier with no account, and EditableImage read every word outside the table correctly. The difference is what happens to everything that is not a word: 2pptx turned ticks into letters and the logo into text, where EditableImage keeps both as exact pictures. Either way, run the five checks at the end of this guide before the slide is presented.
What screenshot gives the best conversion?
One slide per capture at 100% zoom or in full-screen presenter view, saved as PNG, with the window chrome and thumbnail strip cropped out. From a PDF, capture at page width, which is what the test file is. The screenshot-to-PPT page covers each source in turn.