The post title about it being "pixel-faithful" is a bit strange. I don't see that claim in the repo, and they don't seem to even claim full feature support at the moment. And for the features marked as supported in .pptx's, it does seem that at least slide image backgrounds and bullet point images aren't actually working, and some text objects have inverted text colors. Seems quite far away from being pixel-faithful in fact.
Very nice, the rendered demo for all the file types appear to render flawlessly and load instantly on page load, and looking in the DevTools the parsers are split into different Wasm bundles for each file type xslx, docx and pptx:
docx 458kb raw 217kb gzipped
pptx 574kb raw 253kb gzipped
xslx 601kb raw 269kb gzipped
I expected the Wasm bundles to be large and a lot more bigger than that for some reason.
ChatGPT.com can benefit from using this library (or such a library) for rendering a preview of the file in a side panel on the right, instead of just giving me a download link to the outputted/transformed docx/pptx/xslx file.
So I just tried this with a bunch of medium-complex documents and it's wildly wrong. I suspect the authors have never seen an actually complicated Word document?
Interesting because I'm building ooxml-cli right now, for editing pptx, docx, xlsx.
At work I had to adapt a pptx to a corporate template and tried via agent. It kept failing so I started building and then it was able to relatively quick and accurate do what I needed.
Then I needed it to make tables, add pictures. Recently wanted to get data from an xslx and replace text in a presentation etc.
So the tool is growing and maybe this would be interesting to have as the non LibreOffice dependent viewer...
I tried a few PPTX files from consulting firms that are available online. The rendering does not seem to be truly pixel-perfect, but all of them were quite readable and had a good layout, which is already an impressive feat.
I literally had to solve the "preview Office files in the browser" problem last week. I couldn't find a decent solution, so I ended up making a endpoint that ran the files through headless libreoffice on the server to convert them to PDF.
For PPTX and DOCX, this solution is slightly worse than libreoffice conversion (this does not appear to output highlightable text, while PDF conversion does).
However, the XLSX preview BLEW my mind considering this was AI coded. Really good, even interactive!
Pretty cool, rendering PowerPoint files to an image is probably the only way for LLMs to make sense of them.
Does this work in Cloudflare’s workerd environment? Would be nice to have a cheap serverless render -> LLM (GLM-OCR / PaddleOCR) -> Markdown pipeline for the various MS Office formats.
This code creates a JSON intermediate representation that LLMs could probably consume. You might want to simplify it to focus on content and reduce token usage.
If someone actually got "pixel-faithful" Office documents rendering correctly, MS would be screwed. That's actually really important for a lot of companies that carry around decades-old templates that never look exactly right in LibreOffice or any other software that attempted to replicate it.
The slightest misalignment of a paragraph means a line on page 27 of 120 now moved down by 2 pixels, screwing everything else out of alignment. Yes, plenty of companies pay Microsoft 365 subscriptions because of exactly this reason; it sounds ludicrous when you think they could just pay someone to replicate the formatting in a different suite a lot less than the subscription costs, but that's not how it works...
Sadly, Microsoft 365 is not “pixel perfect” compared to word. I often run into headaches where line numbers are different between the two and content ends up on different pages.
If Microsoft can’t get consistent rendering of word docs between Word for Windows, Word for macOS and Office 365, I don’t like anyone else’s chances.
Can the same version of Word now produce the same rendering on two PCs? In the past (I didn't really check recently, I'm thinking more about 10 years ago) the same file might have had page breaks in different positions and things like that. I never understood if it came from slightly different versions of the fonts, from some info derived from the default printing or anything else...
Total "Damn if I do, damn if I don't" situation . I put a similar disclaimer on my AI stuff too. It would be much easier if they didn't mention it. But if they're like me, they want to give people the means for an informed decision. We can respect that and move on. Do we even know if actual big software companies aren't doing it?
which means it probably gets all the halucinated assets correctly and any real world documents wrong.
Still, looks pretty; if it actually has proper testing, could close the gap. Code not being the hard part is a major impediment to good software coming out of these things.
I ran it against some of our internal test files, and it failed on all of them how was this project even tested for proper compliance
The post title about it being "pixel-faithful" is a bit strange. I don't see that claim in the repo, and they don't seem to even claim full feature support at the moment. And for the features marked as supported in .pptx's, it does seem that at least slide image backgrounds and bullet point images aren't actually working, and some text objects have inverted text colors. Seems quite far away from being pixel-faithful in fact.
That phrasing was taken from their website: https://ooxml.silurus.dev/
Very nice, the rendered demo for all the file types appear to render flawlessly and load instantly on page load, and looking in the DevTools the parsers are split into different Wasm bundles for each file type xslx, docx and pptx:
I expected the Wasm bundles to be large and a lot more bigger than that for some reason.ChatGPT.com can benefit from using this library (or such a library) for rendering a preview of the file in a side panel on the right, instead of just giving me a download link to the outputted/transformed docx/pptx/xslx file.
So I just tried this with a bunch of medium-complex documents and it's wildly wrong. I suspect the authors have never seen an actually complicated Word document?
Interesting because I'm building ooxml-cli right now, for editing pptx, docx, xlsx. At work I had to adapt a pptx to a corporate template and tried via agent. It kept failing so I started building and then it was able to relatively quick and accurate do what I needed. Then I needed it to make tables, add pictures. Recently wanted to get data from an xslx and replace text in a presentation etc.
So the tool is growing and maybe this would be interesting to have as the non LibreOffice dependent viewer...
I have a bunch of goodies at https://rcarmo.github.io/projects/go-ooxml and https://rcarmo.github.io/projects/python-office-mcp-server you might enjoy then.
I tried a few PPTX files from consulting firms that are available online. The rendering does not seem to be truly pixel-perfect, but all of them were quite readable and had a good layout, which is already an impressive feat.
I literally had to solve the "preview Office files in the browser" problem last week. I couldn't find a decent solution, so I ended up making a endpoint that ran the files through headless libreoffice on the server to convert them to PDF.
For PPTX and DOCX, this solution is slightly worse than libreoffice conversion (this does not appear to output highlightable text, while PDF conversion does).
However, the XLSX preview BLEW my mind considering this was AI coded. Really good, even interactive!
> ...this does not appear to output highlightable text
Yeah, it does.
https://ooxml.silurus.dev/storybook/?path=/story/docxviewer-...
I can't highlight text in Safari or Firefox on my iPhone (iOS 26.5), at least on that first page.
I'm not familiar with this application, so perhaps I'm missing a step, and editing mode.
Misread that as open office xml not office open xml. I wish the standards were named more differently. They are too easy to confuse
Microsoft did that deliberately.
Pretty cool, rendering PowerPoint files to an image is probably the only way for LLMs to make sense of them.
Does this work in Cloudflare’s workerd environment? Would be nice to have a cheap serverless render -> LLM (GLM-OCR / PaddleOCR) -> Markdown pipeline for the various MS Office formats.
This code creates a JSON intermediate representation that LLMs could probably consume. You might want to simplify it to focus on content and reduce token usage.
If someone actually got "pixel-faithful" Office documents rendering correctly, MS would be screwed. That's actually really important for a lot of companies that carry around decades-old templates that never look exactly right in LibreOffice or any other software that attempted to replicate it.
The slightest misalignment of a paragraph means a line on page 27 of 120 now moved down by 2 pixels, screwing everything else out of alignment. Yes, plenty of companies pay Microsoft 365 subscriptions because of exactly this reason; it sounds ludicrous when you think they could just pay someone to replicate the formatting in a different suite a lot less than the subscription costs, but that's not how it works...
Sadly, Microsoft 365 is not “pixel perfect” compared to word. I often run into headaches where line numbers are different between the two and content ends up on different pages.
If Microsoft can’t get consistent rendering of word docs between Word for Windows, Word for macOS and Office 365, I don’t like anyone else’s chances.
Can the same version of Word now produce the same rendering on two PCs? In the past (I didn't really check recently, I'm thinking more about 10 years ago) the same file might have had page breaks in different positions and things like that. I never understood if it came from slightly different versions of the fonts, from some info derived from the default printing or anything else...
I've heard that the same file can render differently in MS Word across different machines and OSs. So, that won’t help either.
Its kind of sad that the first thing in the repo is a mention that no human was involved in the programming.
Total "Damn if I do, damn if I don't" situation . I put a similar disclaimer on my AI stuff too. It would be much easier if they didn't mention it. But if they're like me, they want to give people the means for an informed decision. We can respect that and move on. Do we even know if actual big software companies aren't doing it?
Would author be able to do it otherwise? Is particular tool choice making result worse?
Bit identical/pixel-faithful reproductions are easy to verify…
"LLMs are amazing, I'm so much more productive now"
"oh yeah? Show me what you made, you can't, nobody can, it's all just AI psychosis"
"I made a pixel perfect Office document viewer"
"well... I wish you hadn't"
“If you use LLMs, you’re not a real developer, you’re lazy.”
The best developers are lazy.
Obligatory: https://xkcd.com/378/
Would this project exist otherwise? i doubt it
I'm fine with that, even as someone who hates AI.
which means it probably gets all the halucinated assets correctly and any real world documents wrong.
Still, looks pretty; if it actually has proper testing, could close the gap. Code not being the hard part is a major impediment to good software coming out of these things.
> office-open-xml-viewer
"Built entirely by Claude through iterative prompting."
Holy cow!!