Images & photos · Image Converter & Compressor
Why quality 80 in WebP and quality 80 in JPEG are not the same
· How it works
image-formats image-quality webp
Quality sliders look universal, but the number maps to encoder-specific quantisation and means something different in every format. This post explains what a quality value actually sets, why 80 is not 80, and how to compare formats honestly.
Two files, same slider, different results — the mismatch developers hit when they swap formats but keep the number
Selecting 80 for JPEG and 80 for WebP aligns two controls, not two visual outcomes. ToolAcre passes the same normalized number to different browser encoder modes, and each format produces its own bytes. A developer who swaps the extension while preserving the slider has changed format and encoder simultaneously, so equal input cannot establish equal damage or equal size.
The panel avoids pretending otherwise by measuring the encoded Blob. Use those actual before-and-after bytes as one result and inspect the image as another. If a default must be chosen for a site workflow, test representative photographs, graphics and transparency rather than treating the shared scale as a cross-format calibration.
What a quality number really controls — a scale factor applied to the encoder's quantisation tables, not a percentage of anything measurable
The workbook calls quality a scale factor over quantisation tables. That level of encoder detail is not implemented or surfaced by this repository. `clampQuality` merely converts the UI percentage into a finite number between 0.01 and 1, and `encodeCanvas` supplies it to the browser. No table, coefficient or retained-information percentage comes back.
This correction prevents a misleading mental model. Eighty does not mean eighty percent of source detail, an eighty percent file-size target or a score comparable across browsers. It is a request parameter whose effect can be learned only from the produced file. ToolAcre’s strongest evidence is therefore output bytes plus visual inspection, not a formula around the slider.
Quality is an opaque browser-encoder input, not a repository-defined table scale
JPEG and WebP have separate MIME types and separate entries in the format table. Both are marked lossy, both accept quality, and WebP retains alpha while JPEG does not. Beyond those product facts, browser-specific mappings remain hidden. The application does not assert identical curves, quantisation tables or perceptual behavior for one integer.
The same browser can also evolve its encoders between releases, while another platform may produce a file a few kilobytes different at the same setting. Record browser and version when stable output matters. If byte-for-byte reproducibility is a requirement, a built-in canvas encoder without pinned implementation details is not the appropriate contract.
JPEG and WebP are separate encoders; their private mappings are not exposed here
Canvas APIs accept quality in a nominal zero-to-one range for lossy encoders. ToolAcre narrows its accepted settings to 0.01 through 1 and starts at 0.82. PNG omits quality entirely because its format record says the control does not apply. These are application behaviors verified by code and tests.
The repository does not establish what default a browser would select if quality were omitted for JPEG or WebP; the plan normally supplies one. It also does not expose encoder effort or tuning flags. A comparison should therefore state the exact value ToolAcre used rather than extrapolate from unspecified browser defaults or another application’s “80” label.
ToolAcre passes a clamped 0.01-to-1 value and does not define browser defaults
Outcome matching begins by choosing the axis that matters. For a strict upload limit, compare files near a similar measured size, then prefer the cleaner one. For an editorial quality floor, lower each format independently until artefacts become unacceptable and compare the resulting bytes. Equal slider positions answer neither question.
Keep dimensions fixed during this stage. Resizing can dominate size and introduce its own resampling changes, making the format comparison hard to interpret. Once an encoder choice is understood, repeat at actual display dimensions. This order distinguishes format behavior from the separate benefit and cost of serving fewer pixels.
Worked example: one photo exported at several qualities in both formats — pairing files of similar size and inspecting them at 100 percent
Choose a source photograph containing skin, foliage, a smooth sky and a sharp building edge. Export JPEG and WebP at 60, 75, 82 and 90 without resizing. ToolAcre will report real output bytes for each run; no article can truthfully pre-fill those values because image content and browser encoder determine them.
Pair outputs whose measured sizes are closest, then inspect the same regions at native scale. Note transparency separately if the source has it, because JPEG will use the chosen matte while WebP can keep alpha. The comparison is complete only when size, visible damage and required format features are all recorded.
What this does not cover — perceptual metrics such as SSIM, encoder tuning flags and AVIF or JPEG XL
ToolAcre does not compute SSIM or another perceptual metric, expose advanced encoder flags, or offer AVIF and JPEG XL output. Its supported encoders are PNG, JPEG and WebP. The absence of a metric means the editor remains responsible for the visual acceptance threshold instead of receiving a numeric verdict that the product never calculated.
It also has no target-size search. You choose quality and observe the outcome; the tool does not iterate until a named byte ceiling is reached. For a hard limit, make deliberate passes from the original and stop after a satisfactory result. Recompressing the previous output turns the search into generation loss.
Takeaway: compare results, not numbers — how the Image Converter & Compressor lets you try both formats on the same image with nothing uploaded
Compare results, not labels. A quality value becomes meaningful only alongside target MIME type, browser, dimensions, measured bytes and a visual check. This richer record also explains why a later browser or different image may need a different number without implying that either encoder is broken.
The Image Converter & Compressor supports the experiment locally: one source, independently chosen JPEG and WebP settings, honest lossy warnings and measured downloads. Preserve the master and select the smallest result that meets the actual visual and compatibility requirement, even when its slider number differs from the other format.