Images & photos · Image Converter & Compressor
16-bit PNG and 8-bit output: what bit depth means for your images
· Background
image-formats bit-depth canvas
Bit depth sets how many distinct levels each colour channel can hold: 256 at 8 bits, 65,536 at 16. This post explains where 16-bit files come from, why browsers work at 8 bits, and when the reduction matters.
The 16-bit scan that halved in size after conversion — the clue that something more than compression happened
A 16-bit scan that becomes much smaller after conversion may have changed format, metadata, profile, dimensions, compression or sample precision. File size alone cannot isolate which one. ToolAcre does not report input or output bit depth, so this symptom should trigger inspection with a format-aware tool rather than an immediate conclusion.
Protect the scan before testing. Browser conversion produces a delivery derivative and should not overwrite the only high-quality source. Record the source’s verified properties, then hold dimensions constant for the first export. A controlled comparison is the only way to separate expected delivery changes from accidental workflow loss.
File-size change alone cannot prove a bit-depth reduction
Bit depth describes how many binary values can represent a channel: eight bits provide 256 possible codes and sixteen provide 65,536. That arithmetic is general, but it does not tell you what a particular file stores or what a browser preserves. Container inspection must establish the source, and output inspection must establish the derivative.
More codes can support smoother gradients and heavier tonal edits before gaps become visible. They do not guarantee a better photograph, wider gamut or accurate display. Sensor data, colour profile, transfer curve and editing path are separate properties. Avoid reducing all image quality to one bit-depth label.
Exact level counts are general binary arithmetic, but source and output depth still require file inspection
The workbook assigns specific channel depths to PNG, baseline JPEG and WebP. Those specification claims are not encoded or tested in ToolAcre, whose format table tracks only lossiness, alpha and whether quality applies. This module therefore omits a format matrix that would require external standards evidence.
The application can still make a safe product statement: canvas re-encoding does not promise preservation of source bit depth. A successful PNG output means the available PNG encoder stored its supplied canvas pixels losslessly; it does not mean a high-bit source remained high-bit through decode, draw and encode.
Format bit-depth matrices are omitted without authoritative specifications
The outline states that browser pipelines are universally 8-bit. The inspected renderer requests a two-dimensional context and draws an ImageBitmap, but it never reads pixel arrays, selects a pixel format or asserts channel depth. Browser implementations can evolve, so this article does not hard-code an internal representation the product cannot observe.
Treat the canvas as a precision boundary of unspecified implementation detail. If retaining a known high bit depth matters, use a format-aware library or desktop workflow that exposes that property and verifies the output. ToolAcre is designed for common browser delivery, not high-precision master preservation.
The repository does not expose or test canvas channel depth, so no universal 8-bit claim is made
A derivative may look fine for ordinary viewing and fail after strong tonal adjustments because a delivery workflow retained less editing latitude. The repository has no banding detector or edit simulation, so “harmless” cannot be generalized. Judge the actual downstream use: final display and future editing are different requirements.
For a final web gallery, compare gradients and shadow transitions at intended size. For a file expected to undergo correction later, keep the richer source and postpone browser conversion until delivery. This policy does not depend on proving exactly how one browser stores its canvas; it follows from preserving options at the master boundary.
Worked example: a 16-bit PNG scan to WebP for a web gallery — following the levels through the conversion while the 16-bit original stays untouched
Take a verified high-bit-depth PNG scan, duplicate it and convert the copy to the required web format without resizing. Inspect the derivative’s format and channel depth with an independent analyzer, then compare visible gradients in the target browser. Apply any planned display adjustments to a test copy and check for banding.
No fixed size reduction or channel count is claimed in advance. The method produces evidence for the chosen browser and file. Keep the original alongside conversion notes. If output precision fails the requirement, stop and select a pipeline with explicit high-bit support rather than iterating quality on an unsuitable canvas path.
Worked method preserves the source and verifies the derivative instead of claiming an internal level count
HDR formats, camera RAW and floating-point editing workflows are outside accepted inputs. ToolAcre rejects HEIC, HEIF, AVIF, TIFF and RAW under its documented compatibility policy. It cannot replace a scanner archive application, raw developer or high-dynamic-range finishing system.
Nor does a PNG destination alone solve the problem. Lossless encoding preserves the canvas pixels it receives, not source values that may have been transformed earlier. Precision requirements must cover the entire chain from decode through render to file, not only the extension chosen at the end.
Takeaway: 8 bits for delivery, more for masters — how the Image Converter & Compressor produces 8-bit outputs and why you should keep the original
Use browser conversion for delivery copies only after verifying that the result meets the real display need. Keep high-bit-depth sources for future editing and archival decisions. This advice remains sound without overstating a universal canvas channel depth or format specification.
ToolAcre offers a convenient local derivative path with visible format and geometry decisions. Its honesty lies in what it does not promise: metadata, profiles and source precision are not guaranteed. Preserve the master, inspect the output and choose a specialized pipeline whenever those properties are contractual.