Adam7 stores PNG samples in seven passes so a decoder may show a rough full image before every sample arrives. It does not automatically improve final pixel quality, so measure progressive display, file size, and converted output in the real delivery environment.
PNG → WebP
Understand what Adam7 changes
PNG interlace method 1 is known as Adam7. The specification defines seven distinct passes over the image. Early passes contain samples from widely spaced positions, and later passes fill the gaps until the complete raster is available. A decoder that renders partial data can therefore show a rough version of the whole picture and refine it as more bytes arrive. Adam7 is not a stack of separately authored resolutions. It is a prescribed order for serializing samples that belong to the same final image.
Once the entire file is received and correctly decoded, an interlaced and a non-interlaced PNG can represent the same dimensions and sample values. The Adam7 setting does not increase resolution, recover detail, or make colors inherently more accurate. Removing interlacing does not by itself reduce the final image quality either. Separate three questions before testing: whether an early preview is useful, whether the encoding affects file size or processing, and whether a later format conversion changes the decoded pixels. Each question requires a different measurement.
- Seven passes progressively fill sample positions
- Final resolution is not increased
- Preview behavior and final quality are separate questions
Adam7 changes the order in which samples arrive, not the intended final picture.
Decide whether progressive display has value
The early full-frame impression can help when a large PNG is loaded over a slow connection and the viewer paints intermediate passes. It may be less useful for a tiny icon, a file loaded instantly from cache, or an application that waits before presenting decoded output. Browser rendering, network buffering, page layout, content delivery, and caching can all determine whether a user sees the intermediate stages. Test the actual page with a throttled connection and an empty cache instead of assuming that the file flag guarantees a visible experience.
Interlacing is only one part of delivery performance. Supplying dimensions appropriate for the display, selecting a suitable format, compressing efficiently, setting cache headers, and prioritizing important images may matter more than a progressive preview. Classify the asset: a large hero graphic, a small interface control, a downloadable master, and an offscreen illustration have different requirements. A preservation copy and a web derivative do not need identical settings. Use Adam7 only when its behavior supports the experience of the specific asset and destination.
- Test with throttling and an empty cache
- Distinguish large visible images from small assets
- Evaluate interlacing alongside delivery optimization
Measure size under controlled conditions
PNG filtering and compression operate on scanlines, and the specification treats each interlace pass as an independent image for filtering purposes. As a result, interlacing can change the compressed size even when the final pixels are identical. The direction and magnitude of that difference depend on content and encoder choices, so a universal percentage is not reliable. Create interlaced and non-interlaced versions from the same source with the same encoder, then test photographic content, flat graphics, narrow lines, indexed color, and transparent areas separately.
Keep dimensions, color type, bit depth, palette choices, and metadata equal during the comparison. If one file also removes metadata or changes from indexed color to truecolor, the result no longer isolates Adam7. Record bytes, encoding time, decoding behavior, and the time to a useful first display. If the size and experience differences are negligible, a simpler workflow may be preferable. Save the encoder version and complete settings with the measurements so a future tool update can be tested against the same baseline instead of an undocumented memory.
- Keep dimensions, color type, depth, and metadata equal
- Use representative content classes
- Record encoder version and full settings
Check pixels after conversion to another format
When PNG is converted to WebP or another format, the converter normally decodes the completed PNG raster and encodes that raster in the destination format. The seven-pass storage pattern is not a visual feature that must be copied into the new file. Focus instead on whether the destination uses lossy or lossless encoding, preserves alpha correctly, maintains dimensions, and handles color information as expected. Compare the source and output at the same zoom, paying special attention to thin lines, small text, gradients, saturated edges, and partially transparent pixels.
A lossy destination may add ringing, block-like change, or softened edges. A lossless destination may preserve pixels while still omitting metadata or changing color interpretation. If the source contains animation or an unusual structure, confirm which image or frame was selected. Successful opening is only the first check. Reopen the download in the editing application and the publishing environment that will consume it. Keep the PNG original so a failed derivative can be recreated from a known source rather than being repeatedly transcoded.
- Compare fine edges, text, gradients, and alpha
- Confirm dimensions and selected frame
- Keep the original PNG available
Validate the decision on a representative page
Produce an interlaced and a non-interlaced PNG from one controlled source. Record their pixel dimensions, byte sizes, and relevant file properties. Clear the browser cache, load each version under the same throttled network profile, and observe the time to the first useful impression and the complete image. Then make the required converted derivative and compare it at the actual CSS or application display size. A large inspection zoom can reveal defects, but the page-level experience determines whether progressive display provides practical value.
Keep Adam7 if the intermediate rendering is genuinely useful and its size and processing tradeoffs are acceptable. Choose non-interlaced output if the intermediate stages never appear, the file grows without benefit, or downstream tooling becomes unreliable. Repeat the test with at least one transparent graphic, one photograph-like PNG, and one small interface asset before setting a batch policy. After an encoder, browser, or delivery change, rerun a small sample because the environment, not just the source file, controls the observed result.
- Compare under the same network and cache conditions
- Inspect both actual display size and zoomed detail
- Set policy by asset class rather than one global preset
The value of progressive display must be demonstrated in the real delivery path.
Key takeaways
- Adam7 arranges PNG samples in seven passes so a decoder can build a progressively more detailed full image.
- Interlacing changes sample order, not the final decoded dimensions or inherent color accuracy.
- File-size and early-display effects depend on image content, encoder settings, decoder behavior, and delivery conditions.
- After conversion, compare dimensions, alpha, fine edges, text, gradients, and actual page loading rather than the label alone.
Frequently asked questions
Does Adam7 make a PNG sharper?
No. It changes sample ordering across seven passes and does not raise the final decoded resolution.
Is an interlaced PNG always larger?
Not necessarily. The difference depends on image content, filtering, compression, and encoder behavior, so controlled measurement is required.
Will every browser visibly show the seven stages?
Not always. Network buffering, cache, decoder, and rendering behavior determine whether intermediate passes are noticeable.
Does Adam7 carry over when PNG becomes WebP?
The converter usually decodes the completed raster and encodes new output; the PNG pass structure itself is not transferred as a visual feature.
What should I compare before batch conversion?
Compare size, first and complete display timing, dimensions, text, edges, gradients, alpha, and the result on the real page.