Use an explicit 1 MB budget and inspect the actual result
This target-size workspace uses 1,000,000 bytes as its ceiling. The label 1 MB is decimal, not a binary KiB or MiB measurement. It is a chosen working budget rather than a universal platform rule. It can be useful for larger photographs where detail deserves a greater byte budget, but content, dimensions and visual detail still determine whether a suitable result fits. A larger budget does not guarantee lossless output or require an already-small source to be re-encoded.
The engine keeps the requested dimensions while trying a bounded set of WebP or JPG quality settings. It checks actual encoded byte lengths rather than displaying an estimated success. The current search permits at most twelve attempts and does not assume size always changes monotonically with quality. If no tested result meets the ceiling, the interface marks it as over target. Reducing both dimensions to 75 percent requires your confirmation; it is not a hidden step.
When the source already meets the ceiling, its format matches the selected output and displayed dimensions remain unchanged, the engine may offer the original after actually decoding it. That shortcut retains existing metadata and is explicitly identified. Use the separate metadata-removal workspace when privacy requires a fresh encoding. A large pixel count can still cause memory or time limits even when the compressed file is under one million bytes.
Worked example
A generated file of 990,000 bytes is below this 1,000,000-byte ceiling; a file of 1,010,000 bytes is above it. These are arithmetic examples, not promised results for a photograph. Check the downloaded file length and visual detail. If the current dimensions cannot produce an acceptable result, choose a smaller explicit size or another allowed byte budget.
Local processing and limits
One supported static image is processed at a time. The general input limit is 10 MiB and twelve million pixels; AVIF processing is limited to four million pixels. Unsupported animation, image sequences, HDR and unverified container features are refused. Large images need more memory even when their compressed file is small. A failed operation does not switch to a server-upload fallback.
The browser downloads its own application and codec resources from the same host. Your selected image, result, filename and metadata are not sent to an upload service. Images are not written to localStorage, IndexedDB or a service-worker recovery cache. Reloading the page leaves the image workspace empty. Downloaded files remain under your control; inspect them before sharing.