Does target-size mode guarantee the exact requested size?
No. It searches for the highest tested JPEG or WebP quality at or below the requested ceiling. Some sources cannot reach that ceiling within the supported quality range.
Last reviewed: July 23, 2026
Compression methodology
BatchMint uses a fixed local order: identify, correct orientation, resize, watermark, encode, rename and export. JPEG and WebP can run a bounded target-size search, but no setting guarantees an exact byte result or identical quality across codecs.
The fixed local pipeline
| Step | What happens | Why the order matters |
|---|---|---|
| 1. Identify | Validate JPEG, PNG or static WebP input | Unsupported or damaged files fail before expensive work |
| 2. Orient | Read and correct JPEG EXIF orientation | Resize and watermark coordinates use the visible orientation |
| 3. Resize | Apply maximum width without enlarging a smaller source | Watermark size is calculated against final pixel dimensions |
| 4. Watermark | Composite optional text or a PNG/WebP logo | The mark is included in the pixels passed to the encoder |
| 5. Encode | Create JPEG, PNG, WebP or AVIF output | Format and quality determine the final byte stream |
| 6. Name and export | Render the template and save files, folder, ZIP volumes or CSV | Export operates on completed local Blobs |
Orientation and resizing
JPEG camera orientation is handled before downstream transforms so portrait images are not resized or watermarked with swapped coordinates. Pica performs the local resize while preserving aspect ratio. The maximum-width control never enlarges a smaller source because upscaling adds pixels without restoring missing detail and can make compression artifacts more visible.
Watermark compositing
Text and logo watermarks are applied after resizing and before final encoding. Text supports font size, color, opacity, rotation, shadow and outline. Logo input is a local PNG or WebP Blob with a width expressed as a share of the final image. Nine-grid placement uses a margin; tiled placement uses a staggered, size-adaptive grid and ignores the single-position margin.
Format-specific encoding
| Format | Current encoder path | Transparency | Important behavior |
|---|---|---|---|
| JPEG | jSquash MozJPEG-compatible codec | No | Transparent pixels are composited onto the selected background |
| PNG | PNG encode followed by OxiPNG level 2 | Yes | Useful for lossless graphics; detailed photos can remain large |
| WebP | jSquash libwebp-based codec | Yes | Supports photographic output and alpha; destination compatibility still matters |
| AVIF | On-demand browser codec | Yes | Loaded only when selected; encoding is slower and outside-browser support varies |
How target-size search works
JPEG and WebP target-size mode runs a bounded binary search between configured quality limits. Each candidate is encoded and measured. When one or more candidates are at or below the requested bytes, the engine retains the highest-quality candidate found in that range. If the smallest tested candidate is still too large, the job reports that the target could not be reached instead of pretending an exact match exists.
- The search is bounded so a single file cannot encode indefinitely.
- A target is a ceiling search, not a promise of exact byte equality.
- Quality numbers are encoder-specific and do not represent identical visual quality between JPEG, WebP and AVIF.
- Image dimensions, noise, texture, transparency and watermark pixels all change the reachable result.
Transparency and JPEG background
PNG, WebP and AVIF can preserve alpha transparency. JPEG cannot. Before JPEG encoding, BatchMint composites transparent pixels onto the user-selected background color. This is an actual pixel operation, not metadata, so the chosen color becomes permanent in the JPEG result. Transparent product cutouts should be checked for edge halos against the intended storefront background.
What a quality setting can and cannot tell you
A quality value is a useful starting point within one encoder, but it is not a universal visual scale. Two images exported at quality 82 can have very different byte sizes, and JPEG 82 is not visually equivalent to WebP 82. BatchMint's published benchmark intentionally records exact encoders, fixtures, dimensions and bytes instead of claiming a universal percentage saving.
A repeatable review method
- Keep an untouched source outside the export folder.
- Choose one representative image containing texture, gradients, text and relevant transparency.
- Process it with the final dimensions, watermark and format settings.
- Compare at the actual display size, then inspect 100% view for blocks, halos, banding and blurred labels.
- Record the recipe and test several different images before applying it to the full batch.
Measured evidence and format guidance
- Reproducible compression benchmark
Deterministic fixtures, encoder versions, settings, source files and exact output bytes.
- How to choose JPEG, PNG, WebP or AVIF
A destination-first decision guide for transparency, compatibility and content type.
- Testing methodology
How measured facts are reproduced and where conclusions stop.
FAQ
Where does image encoding run?
Decoding, resizing, watermarking and heavy encoding run in local Web Workers rather than a BatchMint processing server or the page's main UI thread.
What happens to transparency when exporting JPEG?
Transparent pixels are composited onto the selected background color before JPEG encoding because JPEG has no alpha channel.