You got the email: "Figure 2 is below the 300 DPI minimum," or "please resubmit in TIFF," and the source file is nowhere to be found. Maybe a co-author sent you a PNG. Maybe it's a screenshot from a slide. Maybe the original .ai lives on a laptop that graduated two years ago.
This guide is about that exact situation — fixing a figure when you can't just re-export it.
The 60-second fix: upload your PNG/JPG to SciDraw AI Convert, pick 300 DPI (photos) or 600 DPI (line art), let AI super-resolution rebuild the missing pixels, and download a TIFF (LZW), PNG, or PDF at your journal's print width. You get free daily credits to try it, and it works for Nature, Science, Cell, Elsevier, Springer, Wiley, and ACS.
Figure rejected for resolution or format? Drag in the PNG or JPG, choose your DPI, and download a journal-ready TIFF — under 5 minutes, free daily credits.
Which path is yours?
| Your situation | The honest answer |
|---|---|
You still have the source (.py, .R, .ai, .svg) | Re-export from it at 600 DPI — that's always cleanest. Code snippets are further down. |
| You only have a PNG / JPG / screenshot | You can't add real detail by hand — this is exactly what Convert is for. |
| The text or line art looks jagged even after upscaling | The raster is too degraded; rebuild the panel in Scientific Figure Maker. |
| You're not sure which spec failed | Run Figure Checker first — it reports DPI, size, format, and color mode in one screen. |
| You want every spec before you start | Read Nature, Science & Cell Figure Requirements. |
The rest of this guide explains why "just set it to 300 DPI" doesn't work, and how to choose the right fix without guessing.
Why you can't just "change it to 300 DPI"
This is the single most common misunderstanding, and it sends people in circles.
DPI is not a quality dial. It's a ratio — pixels per printed inch. Changing the DPI number in your file's metadata doesn't create pixels; it just tells the printer to pack the existing ones more tightly, which makes the figure smaller, not sharper.
The math is unforgiving:
- An 800-pixel-wide image at 300 DPI prints at 2.67 inches.
- The same image at 600 DPI prints at 1.33 inches — even tinier.
- Nature's single column is 88 mm (3.46 in), so you'd need at least 1,038 px at 300 DPI, or 2,076 px at 600.
So the real question is never "how do I change the DPI?" It's "where do the extra pixels come from?" You have two sources: re-export from the original (if you have it), or reconstruct them with AI super-resolution (if you don't).
How many pixels do you actually need?
required pixels = print width (inches) × target DPI
| Final print width | 300 DPI | 600 DPI | 1000 DPI |
|---|---|---|---|
| 3.5 in (single column) | 1,050 px | 2,100 px | 3,500 px |
| 5.5 in (1.5 column) | 1,650 px | 3,300 px | 5,500 px |
| 7 in (full page) | 2,100 px | 4,200 px | 7,000 px |
If your image already meets the number in the column you need, don't upscale — you only need a format change. If it falls short, you need more pixels.
DPI is pixels per printed inch, not a quality dial — the same pixels packed tighter just print smaller, not sharper.
If you only have a PNG, JPG, or screenshot
This is the case the manual route can't solve, and it's more common than anyone admits — the figure was made by a collaborator, pulled from an old slide deck, or screenshotted from a draft. There's no source to re-export, so traditional resizing (Photoshop's "Bicubic," Preview's resize) just blurs what's already there.
AI super-resolution is different: instead of stretching existing pixels, it reconstructs edges and detail.
| Plain resizing | AI super-resolution |
|---|---|
| Stretches and blurs edges | Reconstructs sharp edges |
| Blocky artifacts when enlarged | Clean, continuous detail |
| Looks worse the bigger you go | Holds up at print size |
The actual steps
- Open SciDraw AI Convert and drop in your PNG or JPG (up to 10 MB).
- Pick the output format your journal asked for — TIFF (LZW) for most submissions, PNG for transparency, PDF for vector-friendly workflows.
- Pick the DPI tier: 300 DPI (~1K) for photos, 600 DPI (~2K) for detailed figures, 1000 DPI (~4K) for line art and posters. If the source is below target, super-resolution kicks in automatically.
- Download, then verify — open Properties (Windows) or Preview → Inspector (Mac) and confirm the resolution.
No source file? Start from the PNG. AI super-resolution rebuilds the pixels journals need, then exports TIFF/PNG/PDF at the right width.
AI super-resolution reconstructs sharp edges and detail instead of stretching existing pixels, so the figure holds up at print size.
One honest caveat: super-resolution invents plausible detail, not real data. It's reliable for diagrams, schematics, and text-on-figure. For micrographs and gels, reviewers can sometimes spot reconstructed texture — if you have the original acquisition file, prefer that. If the AI-upscaled result still looks soft at print size, that's your signal to rebuild the panel instead (next section).
When upscaling won't save it — rebuild instead
Some figures are too far gone. A 400 px screenshot of a labeled diagram won't upscale into crisp 8-pt text; the letterforms were never there to reconstruct. When the content is simple but the file is ruined, it's faster to rebuild than to fight it.
Scientific Figure Maker regenerates the panel — mechanism, workflow, schematic — as a fresh, high-resolution figure with editable, sharp labels, then exports straight to journal specs. Use it when:
- The text is fuzzy and stays fuzzy after upscaling.
- You only have a tiny thumbnail of your own figure.
- The layout needs changing anyway for the revision.
I still have the source file
If you can re-export, do it — this is always the highest-fidelity path, and it costs nothing. Set the size in physical units so it lands on the journal's column width directly.
Python / matplotlib:
plt.savefig(
'figure.tif',
dpi=600,
bbox_inches='tight',
format='tiff',
pil_kwargs={'compression': 'tiff_lzw'
R / ggplot2:
ggsave("figure.tiff", dpi = 600, width = 180, height = 120,
units = "mm", compression = "lzw") # 180 mm = Nature double columnIllustrator / Inkscape: export at 600 DPI, embed (don't link) images, keep text as text. If the figure is a real SVG and you only need a label or color change before exporting, see Edit SVG Scientific Figures.
That's the whole trick: physical width in millimetres, DPI set explicitly, LZW compression for a smaller lossless TIFF.
The five mistakes that cause re-rejection
- Submitting JPEG. Its compression smears halos around text and lines that production teams flag instantly. Use TIFF or PNG.
- Upscaling line art instead of re-exporting it. Graphs from matplotlib/R/Illustrator should be re-exported at 600 DPI from source — never upscaled from a raster.
- Embedding the figure in Word. Word recompresses images. Always upload separate figure files.
- Using a screenshot as final art. Screenshots are 72–96 DPI; reconstructed detail only goes so far.
- Ignoring color mode. A few journals still want CMYK for print — check before you export.
Frequently Asked Questions
How do I convert a PNG to a 300 DPI TIFF for journal submission?
Upload the PNG to SciDraw AI Convert, choose TIFF as the output, and select the 300 DPI preset. If the source is below 300 DPI at your column width, AI super-resolution adds the missing pixels automatically. Download and confirm it reads 300 DPI in Properties (Windows) or Preview → Inspector (Mac).
Why doesn't "just change the DPI to 300" work?
Because DPI is a ratio, not a quality setting. Editing the metadata from 96 to 300 doesn't add pixels — it shrinks the print size. An 800×600 image at 300 DPI prints at 2.67 × 2 in, far below the 88 mm Nature minimum. You need more pixels: re-export from source, or reconstruct them with AI super-resolution.
What DPI does Nature require?
300 DPI for photographs and halftones (gels, blots, micrographs); 600 DPI for line art (graphs, diagrams, schematics). Mixed figures default to 600 to be safe. Format: TIFF (LZW), EPS, or PDF — never JPEG for the final file.
Can I upscale a screenshot to journal grade?
It depends on the source. Screenshots of vector content (PDF text, web pages) upscale cleanly. Screenshots of raster content (photos, micrographs) won't gain genuine detail — AI fills in plausible pixels, but reviewers can notice on micrographs. When you can, re-export from the original application; when you can't, Convert is the best available fix, and rebuilding in Scientific Figure Maker is the fallback for fuzzy text.
Why TIFF over PNG for journals?
TIFF supports CMYK (needed in print production), lossless LZW compression, 16-bit depth, and embedded ICC profiles. PNG is fine for many journals but may be re-converted in production. If the journal lists TIFF as preferred, submit TIFF.
What's the largest figure size before a journal rejects it?
Nature and Wiley cap individual figures at 10 MB; Cell and Science have no hard limit but prefer under ~50 MB. If you're over, switch 16-bit TIFF to 8-bit, or apply LZW compression — typically 30–50% smaller with no quality loss.
How do I check the current DPI of my figure?
Windows: right-click → Properties → Details → "Horizontal/Vertical resolution." Mac: Preview → Tools → Show Inspector → "Image DPI." Either: drop the file into Figure Checker — it reports DPI, dimensions, color mode, and journal fit at once.
Fix it now
- No source file, low DPI or wrong format → Convert PNG/JPG to TIFF/PDF
- Not sure what failed → Check figure DPI and format
- Text still fuzzy after upscaling → Rebuild the figure
- Real SVG, just needs a label/color edit → Edit SVG figures
Related Guides
- Nature, Science & Cell Figure Requirements — exact size, DPI, and format specs
- How to Check Figures for Submission — pre-submission quality checklist
- Publication-Ready Figures in Minutes — AI workflow for research figures
- Graphical Abstract Requirements by Journal — specs for Elsevier, Nature, Cell Press
- Figure Checker Tool — check DPI, format, and colorblind safety
- Image Converter Tool — convert figures to TIFF, PNG, or SVG
- Journal Figure Requirements Reference — filterable, sourced specs for 10 publishers (CSV/JSON download)
- Figure DPI Calculator — work out the pixel width you need before converting
- Figure Accessibility Kit — contrast and colour-vision checks to run alongside the DPI pass
