CompressPictureCompress an image
Every re-encode happens on this device. No file is posted anywhere.

Batch compress pictures

Drop a working set, apply one declared rule, then judge every file by its own before-and-after ledger instead of trusting one total.

Drop up to 100 images. Nothing is re-encoded until you set a quality or a target size.Click, press Enter, or paste from the clipboard. 120 MB a file: JPG, PNG, WebP, GIF, BMP, HEIC; TIFF where the browser reads it.
In —Out —Saved —

Nothing to hand?

Applies to any JPEG you open. This is the encoder's own scale, not a percentage of anything. Keys 1 to 9 jump to 10 through 90.

PNG mechanism

Applies to any PNG you open. Not one pixel changes. The picture is encoded under more than one row filter and the shortest attempt wins — the step a browser's own PNG writer does not take.

Nothing is re-encoded until a file is open and Run is pressed.

A batch is a queue with an account

The useful output is not only a ZIP. Every row states the detected input, chosen output, original bytes, resulting bytes, encoder setting and whether the operation genuinely saved space. Files are processed in sequence so a large folder does not decode all of its full-resolution images into memory at once.

  1. 1Select or drag the photographs as one group. Their signatures are read from the bytes, not inferred from filenames.
  2. 2Choose the compression mechanism. JPEG uses a quality setting; PNG uses lossless re-deflate or a palette reduction when requested.
  3. 3Start the queue and leave it visible. Each row moves independently through opening, running, done or failed.
  4. 4Inspect no-gain and missed-target rows rather than assuming every new file is preferable.
  5. 5Download the useful results together; the archive keeps distinct names when two source extensions would otherwise collide.

What one rule does not mean

A shared quality number is not a shared visual loss or a shared byte saving. A smooth portrait and a frame of foliage respond differently because their spatial detail differs. The ledger exists so the batch remains auditable at file level.

  • An already efficient JPEG may grow after another encode and should be kept unchanged.
  • A transparent input may leave as PNG because JPEG cannot carry its alpha channel.
  • A target-size search can succeed for one frame and hit the format floor for another at the same dimensions.
  • Metadata removal saves different amounts because files carry different blocks.

The queue writes JPEG and PNG, not WebP.

WebP input can be decoded here, but the queue only ever writes the two formats its rows are built around: an opaque WebP is written to JPEG and one carrying alpha is written to PNG, with that format change stated on its row. The single-file WebP route does encode WebP, so a WebP you want to keep as WebP belongs there rather than in this queue.

Questions about the queue

Does every picture get the same quality number?
The chosen rule is shared, but each file is encoded and measured independently. JPEG and PNG use different settings, and a file that becomes larger is marked no gain rather than presented as an improvement.
Will one broken file stop the queue?
No. Each row has its own opening, consent, run and failure state. A corrupt or unsupported file is reported on its row while the remaining files continue.
Why is there a limit of 100 files?
Finished blobs and the ZIP are assembled in this tab. A fixed ceiling keeps a mistaken folder drop from exhausting memory; larger jobs should be divided into runs whose results can be checked.