How to Translate a PowerPoint Deck as a Freelancer Without a Full CAT Tool

A deck lands in your inbox on Thursday afternoon. Forty slides, due Monday, no project manager, no translation memory, no CAT tool set up for this client. If you want to translate PowerPoint as a freelance translator without building a full CAT project around it, the job splits into four problems, and only one of them is translation. You have to find all the text, translate it with enough context to make sense of eight-word bullets, put it back without wrecking the slides, and check the parts that no spell-checker will ever see. We've watched a lot of freelancers lose a Sunday to the third and fourth problems after finishing the second one in three hours.

Why a PowerPoint deck behaves nothing like a Word document

In a DOCX, text flows. Paragraphs follow each other in a defined order, and whatever tool you open it with reads them in that order. A PPTX has no such thing. Every slide is a set of independent objects, and text lives inside whichever ones happen to hold it: title placeholders, body placeholders, free-floating text boxes, table cells, SmartArt nodes, chart axis labels, grouped shapes, and speaker notes. The reading order depends on the order someone dropped those shapes onto the canvas, which usually has nothing to do with how a human reads the slide.

That has two consequences you feel immediately. The first is that text hides. A deck can look fully translated in the editing window and still have untouched strings sitting in a chart legend, a hidden slide, or the slide master where a recurring footer lives. The second is that every string sits in a container with a fixed width and height. In a Word file, a sentence that runs long is a formatting question you can usually ignore. On a slide it's a broken slide.

We saw this play out on a 42-slide sales deck translated from English into German for a software company. The translation itself was clean and the freelancer delivered on time. Nine slides came back from the client with bullets that had slid behind a product screenshot, because German ran roughly 25 to 30 percent longer than the English source and PowerPoint's autofit had quietly shrunk the type on some slides but not others, where the text box had autofit disabled. Nothing was mistranslated. The deck was still unusable for the pitch it was built for.

Anyone who has worked in a CAT tool on a deck knows the tool handles most of this by extracting every text run in the file and forcing you to confirm each one. Without that safety net, you have to do the finding yourself, deliberately, before you start typing.

What freelance translators need before they translate a PowerPoint deck

Ask for the editable .pptx. Not a PDF, not a Google Slides link, not screenshots pasted into a Word file. A PDF of a deck is a dead end: you can OCR it, but you'll be rebuilding the layout from scratch and charging for work that would not have existed if someone had sent the source file. If the client genuinely no longer has the original, that changes the quote, and they need to hear that before you start.

Then settle four questions in the same email, because each one moves the price.

Are speaker notes in scope? This is the single most common scoping miss we see. Notes are invisible in presentation mode and invisible in a quick scroll through the slides, so people forget they exist. On one investor deck we looked at, 34 slides carried about 900 words of on-slide text and just over 2,800 words of speaker notes. The freelancer had quoted on the visible text. The client assumed the notes were included, since they were "in the file."

Are hidden slides in scope? Decks that have been reused for several audiences often carry a backup appendix that's hidden. Sometimes the client wants it, sometimes it's dead weight from two quarters ago.

Which target font, and does it exist on the client's machine? This matters for Cyrillic, Greek, and anything CJK. A deck built in a display font with Latin-only glyph coverage will substitute silently when you paste in Russian or Japanese, and the substituted font will have different metrics, which puts you back in the overflow problem.

What do they want back? Most clients want the translated deck. Some also want a bilingual reference so a colleague who speaks the language can check it. That second file is easy to produce while you work and expensive to reconstruct afterwards.

Run your own word count too. PowerPoint's built-in statistics are unreliable for this, since they miss text in some object types. Extracting the text first, then counting, is the only number you should quote from.

Getting the text out when you don't have a CAT tool

There are three practical routes, and the deck decides which one fits.

Translating in place, directly in PowerPoint, works for short decks. Under about fifteen slides with low repetition, the overhead of extracting and reimporting costs more than it saves. You click into each text frame and overwrite it. The discipline this requires is a systematic sweep: go slide by slide, use the Selection pane to list every object on the slide, and tick them off. The Selection pane is the closest thing PowerPoint gives you to a segment list, and it will show you the grouped shapes and the text box sitting behind an image.

The second route is a two-column source and target table, built by hand in Excel or a Word table. You paste each string into column A, translate in column B, and paste back. It's tedious, but it gives you something a CAT tool gives you and PowerPoint doesn't: everything visible in one place, sortable, so you can spot that the same phrase appears on six divider slides and translate it once. Be careful with PowerPoint's Outline view here. It looks like an export feature, and it only shows text from title and content placeholders. Text boxes, tables, SmartArt and charts do not appear in it at all, so a deck exported through Outline view will look complete and be missing a third of its content.

The third route is a document translation tool that takes the PPTX in and returns a translated PPTX with the structure intact. This is what most freelancers move to once they've done the manual version twice. The relevant question when picking one is not fluency of output. It's whether the tool returns a real editable presentation with the original shapes, or a reconstruction. A reconstruction means you're doing layout work anyway. We wrote about the format side of this in more detail in how to translate a PowerPoint presentation with AI without breaking the slide layout.

Whichever route you pick, watch the deck once in slideshow mode before you translate a word of it. Six minutes. It will save you more than any other six minutes in the job.

Translating slides with context instead of one box at a time

Slide text is compressed to the point of ambiguity, and the ambiguity is usually resolvable only from the surrounding slide. A bullet reading "Increase margin" can be an instruction to the sales team, a claimed benefit of the product, or a line item in a strategy list, and each reading takes a different grammatical shape in most target languages. A header reading "Run rate" means one thing on a finance slide and something else on an operations slide. If you translate the extracted table top to bottom without looking back at the slides, you will get a version of the deck that reads fine as a word list and lands wrong in the room.

Two habits fix most of this. Keep the deck open in a second window while you work, and translate in slide order rather than in whatever order your extraction produced. And decide the register once, at the start, then hold it: bullets in the target language should either all be sentence fragments or all be full sentences. Mixed registers across slides is the thing that makes a translated deck feel translated, more than any individual word choice.

Build a small glossary while you read, before you translate. Twenty to forty terms is enough for a typical corporate deck: product names, feature names, job titles, metric names, and anything the client uses as a slogan. Decks repeat their own vocabulary heavily, and they get read by people who will notice if the same feature has two names on slides 7 and 23. If the client has an existing glossary from previous work, ask for it even if they didn't offer.

One caution, though. Divider and section slides often repeat a phrase that needs to change with context. "Overview" heading a product section and "Overview" heading a financial section may not take the same target word. Treat consistency as your default and let the slide overrule it when it has a reason to.

Putting the text back and fixing what breaks

Reimporting is where the time goes, and it's worth budgeting for it honestly. On a design-heavy deck, expect layout work to take somewhere between 20 and 40 percent of the time the translation itself took. On a plain deck of title-and-bullets slides, it's closer to 10 percent.

Work through the deck in slideshow mode after pasting, not in the editing view. The editing view lies: text that overflows its frame is often still visible there and simply clipped or hidden in presentation mode.

The things that break, in rough order of frequency:

Text overflow, when the target is longer than the source. Your options are to shorten the translation, reduce the font size, enlarge the text box, or turn on shrink-on-overflow autofit. Shortening is usually the right answer and the one clients prefer, because uneven type sizes across slides look sloppier than slightly terser bullets.

Font substitution, when the original font lacks glyphs for the target script. Check this by looking at the deck on a machine that doesn't have the original fonts installed, or by exporting to PDF and looking at the result.

Chart labels that live in an embedded workbook rather than in the slide. This one caught a translator we know on a quarterly business review deck: every axis label and series name looked translated in the slide preview, but three charts still had English category names because those strings sat in the embedded Excel object behind the chart. You reach them by right-clicking the chart and opening the underlying data.

Grouped shapes and SmartArt, where text sits one or two levels down and doesn't get selected by a click.

Hyperlinks whose display text is the URL itself, which should usually stay in the source language even when the surrounding sentence is translated.

Text baked into images. You cannot fix this without the original image files, and you should tell the client rather than silently leaving it in the source language.

A QA pass that fits in half an hour

Run the deck as a slideshow, at full screen, from slide one to the end, in the target language, and don't skip. This catches overflow, substituted fonts, and any slide you missed entirely. It takes about ten minutes for forty slides.

Export to PDF and scan the pages. PDF export renders exactly what the audience will see and will expose clipping the editing view hid from you.

Check every number, date and currency figure against the source. These are the errors clients notice fastest and forgive least, and they're also the errors that pure linguistic review tends to skip because nothing about them looks wrong in the target language. Watch decimal separators and date order in particular for European target languages.

Use Find to check your glossary terms, one term at a time. It's mechanical and it takes four minutes.

Open the speaker notes pane and scroll through it, even if notes were out of scope, to confirm you left them consistently untouched rather than translating three of them by accident.

Finally, compare slide counts and confirm no slide got deleted during editing. This happens more often than it should when you're pasting under time pressure.

What to charge, and what to say about it

Quote the text and the layout as separate lines. Per-word for the translation, hourly or per-slide for the layout and formatting work. Burying the second one inside your per-word rate means you either overcharge on clean decks or lose money on messy ones, and it makes it impossible to explain to the client why one 40-slide deck cost twice what another did.

Never quote a deck without opening it. A 40-slide deck can hold 700 words or 6,000, and slide count tells you nothing about either. A related trap is the deck that's mostly screenshots with baked-in text: it looks small, it quotes low, and it generates a conversation you'd rather have had upfront.

Tell the client what you did not touch, in writing, on delivery. Images with embedded text, hidden slides, speaker notes if they were out of scope, and any string you deliberately left in the source language such as product names or URLs. This one paragraph prevents most of the revision requests we see on deck jobs.

Before you type a word on the next PowerPoint job: open the file in slideshow mode once end to end, open the Selection pane on three or four of the busiest slides to see what's actually there, check the speaker notes pane, right-click any chart to see whether its labels live in an embedded workbook, and only then produce your word count and your quote. That sequence takes about fifteen minutes and it decides whether the job runs to plan or eats your weekend.

The approach here works best on corporate decks of up to roughly sixty slides with a single target language. It stops making sense when you're handling the same deck into five languages, or when the client updates the source mid-project and expects you to absorb the change. At that point the overhead of a proper CAT setup, with a translation memory that survives to the next version, starts paying for itself. We covered where that line sits in do you still need a CAT tool in 2026.