How to reduce file size
Photos, video, PDFs and documents each shrink for different reasons. Here is what actually works for each, and where the sensible floor is.
Almost everyone who searches this has hit a limit: an upload form, an email attachment cap, a portal that rejects anything over 2 MB. The answer depends entirely on what kind of file it is, because the reasons a photo is large have nothing to do with the reasons a PDF is.
Images
Most photos are large because of resolution and quality settings, in that order.
A modern phone photo is around twelve megapixels. Displayed on a screen, or attached to a form, almost none of that detail is visible — the receiving website will scale it down to a fraction of its size before anyone sees it. Reducing the dimensions before you upload loses nothing you would have seen.
After that comes compression quality. JPEG lets you trade detail for size on a sliding scale, and the useful range is wider than people expect: dropping from maximum quality to around 70% typically halves the file with no visible difference on screen.
The practical approach is to name the size you need and let a tool search for it, rather than guessing at a quality number and re-checking.
There is a third lever that only applies to some files, and when it applies it dwarfs the other two: the format. A photograph saved as PNG is carrying every pixel exactly, which is the most expensive way to store something the eye cannot check — moving it to JPG routinely removes eighty percent or more, and nothing about the picture looks different at normal viewing size. The reverse is the mistake worth naming, because people make it while trying to help: JPG to PNG is a legitimate conversion when a form demands PNG or an edit is coming, and it will make a photograph three to ten times larger while improving nothing.
The format lever cuts the other way when the file arrived already using a modern codec. An AVIF is small precisely because it compresses harder than JPEG does, so converting it to JPEG to make it openable will make it bigger, often by half again — AVIF to JPG shows the predicted size before it runs for exactly that reason. Do that conversion for compatibility, never for weight.
The floor: below roughly 30 KB, a photograph stops looking like a photograph. If your target is under that, you need a smaller crop rather than harder compression.
Video
Video size is bitrate multiplied by duration, and that formula is the whole story.
Because duration is a multiplier, the single most effective thing you can do is trim. Cutting thirty seconds off a two-minute clip removes a quarter of the file before any re-encoding happens, and loses nothing in the part you keep.
After that, resolution: 4K to 1080p cuts the pixel count by three quarters. For anything watched on a phone or in a chat window, the difference is invisible.
When the destination gives you a hard 8 MB, 10 MB or other ceiling, the local video compressor works backward from that byte count, shows the video and audio bitrate budget, and measures the finished MP4 before calling it under the target. Use the original instead when it already fits; another lossy encode cannot improve it.
The floor: below about 500 kbps at 1080p, motion turns to mush. If the target demands less than that, reduce the resolution rather than the bitrate.
Animated GIFs
These get their own heading because they behave like neither of the categories above. A GIF stores every frame as a complete picture drawn from a palette of at most 256 colors, using a compressor from 1987, so a short clip in GIF form can easily outweigh the video it was made from.
Two answers, and which one you want depends on whether the file has to stay a GIF. If it does — an email that must animate, a field that only accepts images — then squeezing it under a byte limit is the job, and the order the levers get pulled matters: palette first, then dithering, then frames, resolution last. If it does not, the same clip written as MP4 is the larger win by an order of magnitude, because H.264 describes what moved rather than redrawing the whole frame.
The floor: a GIF that resists every setting is usually not an animation at all. It is camera footage in a container built for logos, and no GIF compressor rescues that. Where the original video still exists, the cheaper repair is upstream: rebuild the GIF from the video at a lower frame rate and a narrower slice of time, so the weight never enters the file in the first place.
PDFs
PDFs are containers, so two documents that look like scans can need different treatment. One may store each page as an ordinary JPEG; another may use Flate, JPEG 2000, JBIG2 or monochrome fax compression. Text, fonts, forms and vector drawings follow other rules.
Start with an inspection rather than a promise. The local PDF compressor can re-encode the ordinary RGB JPEG images its browser stack can safely handle, then compares the complete candidate PDF with the original. It offers no download when the whole file did not get smaller. That protects you from a common failure: an image stream shrinks while the rebuilt PDF structure adds back more bytes.
When a scan uses an unsupported image encoding, re-scanning at an appropriate resolution or changing the export settings at the source is safer than turning every page into a new JPEG. Whole-page rasterization destroys searchable and selectable text, can break forms and links, and can make a text-heavy document dramatically larger.
Embedded fonts can also contribute. Subsetting keeps only the glyphs actually used, but that requires a PDF optimization engine that understands font programs. A JPEG-only browser compressor should not claim to perform it.
Office documents
A Word or PowerPoint file is a zip archive. Almost all of its size is pasted images — often screenshots pasted at full resolution and then dragged small on the slide, which keeps every original pixel.
Both Word and PowerPoint have a built-in “compress pictures” function that fixes exactly this, and it is usually the entire answer.
What does not work
Renaming the extension. Changing .png to .jpg changes nothing about the bytes.
The file is still a PNG and now lies about what it is.
Zipping an already-compressed file. JPEG, PNG, MP4 and PDF are already compressed. Zip finds no redundancy left and saves a percent or two, sometimes making the file slightly larger.
Repeated compression. Compressing an already-compressed JPEG degrades it again without saving much, because the first pass already removed what the format could cheaply discard. Always go back to the original.