Enter your target
Type the KB limit into the target-size field and pick JPG or WebP output — the formats that compress well.
Compression guide
Government portals, visa applications, job sites, and marketplaces love strict file-size limits: "max 100KB", "under 50KB", "up to 200KB". FitMyPic takes your target size, tests quality levels until the output lands close to it, and shows you the measured result before you download. All locally — your photo never leaves your device.
Need the tighter limits? We keep a dedicated, measured page for the 50KB target — the tightest one forms ask for.
Common limits
How it works
Type the KB limit into the target-size field and pick JPG or WebP output — the formats that compress well.
FitMyPic tests quality levels until the encoded file lands close to your target — no blind trial and error.
The measured output size and live preview are shown before download. If 100KB needs a smaller crop, resize first and compress again.
Questions
FitMyPic tests quality levels to land as close to your target as possible and always shows the measured output size before you download. Image content affects compression, so results may be slightly above or below the target — you see the real number first.
For photos, JPG and WebP compress far better than PNG. PNG is lossless and best for sharp graphics, text, and transparency — not for hitting small size targets on photos.
Compression trades some fidelity for size. At moderate targets the difference is hard to see; at extreme targets (like 20KB from a large photo) quality loss becomes visible. The live preview shows exactly what you get.
No. Compression runs locally in your browser with the canvas API — your files are never transmitted to a server.
When a form demands "under 100 KB", the instinct is to drag the quality slider down until the number fits. On a large image that produces a file that meets the limit and looks like it was photographed through a window screen, because JPEG at very low quality removes detail from every pixel rather than removing pixels that were never going to be shown.
The order that works is the reverse: reduce the pixel count until the file is close to the target, then use quality only for the last adjustment.
There is a useful arithmetic check available before you touch any slider. A file of a given byte size corresponds to a budget in bits per pixel:
bits per pixel = (target bytes x 8) / (width x height)
| Canvas | Pixels | Target | Bits per pixel | What it implies |
|---|---|---|---|---|
| 4000 x 3000 | 12.0 M | 100 KB | 0.07 | Not achievable in JPEG without visible damage |
| 1600 x 1200 | 1.92 M | 100 KB | 0.42 | Possible but visibly soft in detailed areas |
| 1000 x 750 | 0.75 M | 100 KB | 1.07 | Comfortable for a photograph at moderate quality |
| 800 x 600 | 0.48 M | 100 KB | 1.67 | Good quality with room to spare |
| 600 x 450 | 0.27 M | 100 KB | 3.0 | Essentially transparent for most content |
Anything below roughly 0.5 bits per pixel is a warning sign: the encoder is being asked to discard so much that blocking and smearing become visible. It does not matter which slider position produces it.
Measured on a 1600 x 1200 photo-like test image with libjpeg-turbo:
| Quality | Size | vs q75 | SSIM |
|---|---|---|---|
| q60 | 76.3 KB | 0.54x | 0.883 |
| q70 | 112.3 KB | 0.80x | 0.886 |
| q75 | 141.2 KB | — | 0.888 |
| q80 | 193.5 KB | 1.37x | 0.890 |
| q85 | 270.2 KB | 1.91x | 0.892 |
| q90 | 414.7 KB | 2.94x | 0.893 |
From q75 to q90 the file nearly triples while the structural similarity score moves 0.888 to 0.893 — a change most people cannot see. Chasing a target downward has the same shape in reverse: the last few kilobytes come at a disproportionate visual cost.
If the image contains text — a document scan, a screenshot, a poster with a caption — JPEG is the wrong format for the job regardless of the quality setting. From q40 to q95 our UI screenshot grew 170.9 KB to 752 KB while PSNR stayed flat at 26.7 dB. Send PNG and accept a larger file, or reduce the dimensions further.
Requirement pages usually specify three things, and it is worth reading all of them:
When a minimum dimension and a maximum byte size appear together, they define a window: the image must be big enough to be legible and small enough to be accepted. If no single quality setting fits inside that window, reduce the dimensions toward the minimum and try again, and avoid JPEG for anything with text.
Open the exported file, confirm the dimensions and the byte size, and view it at 100% zoom. A form that rejects your submission after the fact costs more time than the thirty seconds this takes.
The compressor on this page reports the output size while you adjust dimensions and quality, so the budget arithmetic above is visible rather than estimated. Processing happens on your device.
Source: JPEG quality values measured on 22 October 2026 with Pillow 11.0.0 (libjpeg-turbo); raw CSV published at ./assets/bench/study-b-jpeg-quality.csv.
Everything in this guide runs locally in your browser — free, no account, no uploads.
Open the free tools