A submission is a packaging problem: one PDF in the order the journal expects, a page number on every page, a file under the system's limit, and figures that can be pulled out at print resolution. The second half of this page covers the reading side, scanned papers and recorded interviews.
Order matters. Assemble and number first, then compress, because every later edit invalidates the file you tested. Export figures from the uncompressed build, since a 300 DPI render of a compressed page inherits the compression. Keep the source files, because compressing is one-way and the submitted PDF is only a copy. Nothing here edits a manuscript: these tools assemble, renumber, render and read files, they do not write them. Apart from interview transcription, every step runs locally, which matters before publication. That distinction decides which files are safe to upload.
1. Merge the manuscript with its supplements
Merge PDF assembles the main text, the figure legends, the supplementary tables and the response letter into one file. The cards can be reordered, pages are copied as they are, and links and bookmarks survive, so a table of contents still works after the merge. Build the file in the order the submission system asks for, and check whether the system wants the supplements in the same upload or a separate one before you merge everything together.
One blocker: an encrypted PDF cannot be read by the merger, so a signed cover sheet or a co-author's password protected draft has to be unlocked first. Merging is also a one-way assembly, so keep the chapter files as the sources in case the order changes after a reviewer comment.
Merge PDF
Join the manuscript, supplements and cover letter in the order you set.
2. Number every page
Add Page Numbers draws a continuous number on each page as selectable text, so it does not flatten the PDF. Six positions, four fonts and a size between 8 and 24 pt cover the usual journal requirement, and the starting number plus a skip-first-page switch handles a title page that should not be counted. Number after merging, not before: numbering each file separately gives every section its own page 1.
There is no prefix field, so a running head or a manuscript ID cannot be stamped onto the number itself. If the journal asks for line numbers or a specific footer format, that has to come from the word processor before export, because nothing here edits the page content.
Add Page Numbers
Continuous numbering with a starting value and a skipped title page.
3. Fit the submission limit
Compress PDF rewrites the internal structure with qpdf: objects are packed into streams, uncompressed streams are compressed, and the cross-reference table is rebuilt. Text and links survive because nothing is re-encoded. That gives a real saving on a PDF exported loosely from a word processor, and almost nothing on a file that was already produced efficiently.
Two limits decide whether it will help. The tool is designed for files under 50 MB, and it does not re-encode the images inside, so a manuscript built from 300 DPI scans barely moves. There is no target size field, so you run it and read the output number. When the file is still too big, compress the images before they enter the PDF, and move the supplements to a second upload if the system allows one.
Compress PDF
Lossless structural rewriting for a smaller file, with a 50 MB input limit.
4. Export figures at 300 DPI
PDF to JPG renders pages at 72, 150 or 300 DPI and writes JPG or PNG. Use the page range to export only the figure pages, for example 4-6, instead of rendering the whole manuscript. 300 DPI is the usual print requirement, and a single letter page at that density is already above eight megapixels, which is why the renderer refuses a page over 40 million pixels.
Understand what the output is: an image. Axis labels, legends and panel letters become pixels, so they are no longer selectable, searchable or editable, and the figure is only as sharp as the embedded image in the PDF. Rendering at 300 DPI adds pixels, not detail. Export from the uncompressed build, because a compressed page carries its compression into the render, and keep the PDF as the source of record.
PDF to JPG
Render chosen pages at 72, 150 or 300 DPI as JPG or PNG.
5. Prepare the image files
If the journal wants the figures as separate image files, Image Compressor handles the whole set at once with a quality slider and a percentage resize, and Image Converter normalises a mixed folder to JPG, PNG, WebP or AVIF. AI Image Upscaler can rebuild edge contrast on a soft raster at 2x, 4x, 6x or 8x, locally, up to 20 images per batch.
The limits here are strict and worth knowing before submission. TIFF, which many journals request for line art, cannot be opened or written by any tool on this site, so a TIFF requirement has to be met elsewhere. HEIC cannot be opened either. Upscaling does not recover detail: it cleans artifacts and sharpens edges, and a figure that was captured at low resolution will look cleaner, not more precise. Output is capped at 8192 px on the long edge and 24 megapixels in total.
Image Compressor
Batch compress the figure set with a quality slider and a live preview.
6. Read the literature and the interviews
PDF to Text handles both kinds of reading. A modern paper carries a real text layer, which is extracted exactly and instantly; an older scanned article is rendered and recognised with OCR instead, and the result panel says how many pages came from each route so you know where to proofread. English is built in and eleven more languages can be selected, and everything runs locally, which matters for an unpublished manuscript under review.
Interviews are the one step that uploads. Video to Subtitles accepts a file up to 30 MB, sends it to a server, and recognises Simplified Chinese and English only, returning SRT, VTT or text with timecodes and speaker labels. A free run covers 2 minutes a day and a free account receives 5 credits daily. For human-participant research that upload is an ethics decision, not just a technical one: check the consent form and the ethics approval, and de-identify the recording first. There is no local speech recognition on this site.
PDF to Text
Extract a text layer directly and run OCR on scanned pages, locally.
What this chain cannot do
- No editing of the manuscript. Nothing here changes a sentence, a figure caption or a reference list.
- No PDF to Word. There is no way back to an editable document; keep the source file.
- No TIFF. Many journals ask for TIFF line art, and no tool here reads or writes it.
- No structural OCR. A table becomes lines, not cells, and two-column text interleaves.
- Handwriting is not reliable. Printed OCR covers eleven languages; a handwritten notebook does not.
- Compression does not shrink scans. The PDF rewrite does not re-encode images, so image-heavy files barely move.
- Upscaling adds no detail. A 2x pass cleans edges; it does not recover resolution the figure never had.
- Transcription uploads the recording. 30 MB, Chinese and English only, and an ethics question for human participants.
- No reference manager, no plagiarism check. This is a file packaging chain, not research software.