Updated 2026-10-24 · Export results below were produced and inspected on this date
A thread is several posts pretending to be one document. Getting a usable PDF out of it is mostly a question of where you start: paste the wrong end of the thread and you get a clean, well-formatted file that is quietly missing half the writing.
The short version
- Paste the last post of the thread, not the first. In two tests on 2026-10-20 the first-post link returned only that one post; the final-reply link returned the chain.
- Five buttons, one conversion: Word (.docx), PDF, Markdown (.zip), Plain text (.txt) and Everything (.zip), the last being all four outputs plus the original images. Markdown is not a lone
.md file — it arrives as a ZIP with the text, a source record and the images.
- Images come with the file. On the test thread the image was embedded in the Word and PDF files and copied at original byte size into the Markdown ZIP.
- There is no PNG or image export. If you need a picture of the thread, see saving a thread as an image for the export-then-capture route.
Start from the end of the thread
Most guides tell you to copy the link of the first post. When we tested that on 2026-10-20, it did not reconstruct the thread — it returned the single post you pasted. The chain is rebuilt by walking backwards from the post you give it, so the link that carries the whole conversation is the final reply.
Two public threads by the same author, parsed on 2026-10-20 through this site's own parser. Character counts are what the parser reported.
| Link pasted | Result | Posts | Characters |
| Thread A — first post | single post | 1 | 891 |
| Thread A — last reply | thread | 2 | 3,672 |
| Thread B — first post | single post | 1 | 381 |
| Thread B — last reply | thread | 2 | 848 |
Practical rule: open the thread, scroll to the bottom, copy the link of the last post. Then check that the file contains as many posts as the preview said it found. If the number is lower than what you can see on X, the chain was cut somewhere — paste the last post again rather than editing the file by hand.
What you get in each format
Thread A, exported four ways on the same day. Same input, same parser, four different containers:
Thread A: 2 posts (2026-10-13 and 2026-10-19), 3,672 characters, 1 image. Exported and inspected 2026-10-20.
| Format | What arrives | Size | Image | Structure |
| PDF | one .pdf | 548,977 B 3 pages | 1 embedded | Generated document, not a page capture: text is drawn as text with embedded fonts, and a footer carries the export note. |
| Word | one .docx | 443,250 B | 1 embedded
word/media/image1.jpg | Editable body text, header and footer, image in its original position. The stored image is 418,440 B — the same bytes as the source file. |
| Markdown | a .zip | 405,019 B | 1 file in
images/ | Three parts: the .md (4,334 B) with YAML front matter, SOURCE.txt (519 B) with author and export record, and the image folder. |
| TXT | one .txt | 1,225 B | placeholder only | Plain text with a header block, per-post separators (1/2, 2/2 with handle and date), [image: 01.jpg] markers and a closing attribution line. No images inside. TXT also drops quoted posts. In the same 2026-10-20 sample the thread ran 3,672 characters but the .txt held 1,212 — the two quoted blocks (286 and 2,624 characters) were not written at all. If the thread quotes other posts, export DOCX or PDF instead; TXT is for plain reading, not for keeping a record. |
For a single long post with four images the same four exports behaved the same way: the Word file carried four images in word/media, the PDF ran to five pages with four images, the TXT kept four [image: …] placeholders, and the Markdown ZIP held four JPEGs plus the .md.
How to check your own file
You do not have to trust a preview. Every claim above is checkable in a minute on the downloaded file:
1
Word: look inside the package.Rename .docx to .zip and open it. Images are in word/media/. If that folder is empty, the images did not come through.
2
PDF: count pages and images.Three pages for a two-post thread with one image is normal. A PDF with the right page count but no images means the images failed to fetch, not that they do not exist.
3
Markdown: unzip before you file it.Inside you should see the .md, SOURCE.txt and images/. Notes apps that import a .md on its own will drop the images — import the extracted folder instead.
4
Any format: read the end.The last post is the one that goes missing when a chain is cut. Compare the final lines with what you see on X.
Which format for which job
- PDF — sending it to someone, printing it, archiving something that should look the same everywhere. It is the least editable of the four.
- Word — quoting, excerpting, translating, or dropping the text into a report. This is the one to pick if you plan to change a single word.
- Markdown — a notes app, wiki, or repository. The YAML block keeps title, author, publish date, export date and source link attached to the text, which matters more than formatting a year from now.
- TXT — the format that will still open in twenty years, and the easiest to grep. You lose the images but keep the order and the source line.
All four come from the same parse; you do not need to run the thread through the tool twice to get a second format. In the tool's format list they appear as Word (.docx), PDF (.pdf), Markdown (.zip), Plain text (.txt) (the plain-text file keeps the prose and the links but drops quoted blocks: in our 20 October 2026 sample a 3,672-character thread arrived as 1,212 characters) (the plain-text file keeps the prose and the links but drops quoted blocks: in our 20 October 2026 sample a 3,672-character thread arrived as 1,212 characters) and Everything (.zip), which bundles the others with the original images. Parsed documents and their exports stay available for about three hours, so download what you need in the same session.
Thread, long post, or X Article — three different pages
These get mixed up, including in the guides that rank for them:
- A thread is one author replying to themselves. The posts are separate objects joined by reply links, which is why the link you paste decides how much you get.
- A long post is one post that happens to be long. Pasted alone it produces one document, and that is correct — there is nothing else to collect.
- An X Article is the long-form editor, not a post. It has its own URL shape and is covered in saving an X Article as a PDF.
The failure mode is always the same shape: a clean file that is shorter than the thing you were reading. Checking the post count against the preview catches it before you need the file.
Limits worth knowing before you start
- No image or PNG export. Word, PDF, TXT, Markdown and a bundle ZIP are the outputs; a picture of the thread is a separate step.
- TXT drops quoted posts. Measured 21 October 2026: a thread whose middle post quoted another tweet exported to TXT with 2,910 of the 3,672 characters the same thread produced as a PDF — the quoted block was missing and no warning was shown. Use Word, PDF or the Markdown ZIP when a thread contains quotes.
- Video is not embedded. A post whose media is a video keeps the link, not the file.
- Public only. Protected accounts, deleted posts and anything behind a paywall cannot be read. If you cannot see the post while logged out, no converter can.
- The chain only grows backwards. Later replies added after the post you pasted are not part of that parse — go back to the current last post if the thread has continued.
- Three-hour window. The parsed copy and its exports are temporary. Keep the downloaded file; do not keep the tab.
What was tested, and what was not
Tested 2026-10-20: two public threads (2 posts each) and one public long post with four images, sent through this site's parser and exported to all four formats from a server-side session. Reported numbers are the parser's own counts and the actual file sizes and page counts of the downloaded files.
Not tested: mobile browsers (Safari on iOS, Chrome on Android), threads longer than two posts, posts whose media is video, protected or deleted posts, and how other tools handle the same samples. No claim here is based on a test that was not run, and no success rate is claimed.
Frequently asked questions
Which post should I paste to get the whole thread?
The last post of the thread, not the first. In two tests on 2026-10-20 the link to the first post returned only that post (381 and 891 characters), while the link to the final reply returned the chain (2 posts, 848 and 3,672 characters).
What does the PDF contain?
A generated document rather than a screenshot of the page: on the 2-post test thread the PDF was 3 pages with the thread text drawn as text and the one image embedded. It carries the author handle, the post dates and the source link.
Is the Markdown a single .md file?
No. Markdown arrives as a ZIP containing the .md file, a SOURCE.txt record and an images folder holding the same image files at their original byte size.
Can I get a PNG or an image of the thread?
There is no image export. The format list is Word (.docx), PDF (.pdf), Markdown (.zip), Plain text (.txt) and Everything (.zip). If you need an image, export first and capture the result — see the image guide.
Do I need an X account, and are protected posts supported?
Public posts need no account. Protected, deleted or paywalled posts are not supported: if a post cannot be read logged out, this tool cannot read it either.
Turn a thread into something you can keep
Paste the last post of an X thread and get a Word, PDF, TXT or Markdown file with the images in place.
Open TweetArticleSave
Related reading on our own editorial blog (we build these tools): What Each X Article Export Format Actually Keeps (and What It Drops) — background for the points above, not a third-party endorsement.
Compare one document in three formats
We sent the same original reading checklist through the production format renderers: PDF, Word and Markdown ZIP. Look for the opening and closing markers. Use the PDF to inspect the page layout, Word to edit the text, and the extracted Markdown file to check the link target. This example tests rendering; a complete example file cannot tell you whether every reply in a live X thread was collected.