An image can fit its box while losing the detail the page needs.
A narrow crop might show one object from a comparison and hide the other. Moving the focal point can change which object survives. It cannot always preserve both.
Start with the image's purpose, its essential detail and the spaces it needs to occupy. This exercise produces a crop brief for the person preparing assets and implementing the page. It does not edit an image, set a CFAI control or establish a tested layout.
Name what the image must communicate
Record the source file and version, its intended placement and the information it supplies. “Hero photo” names a location. “Shows the two sample layouts side by side” identifies meaning that a crop could remove.
Identify the parts that must remain together. That may be a person and the object they are demonstrating, both sides of a comparison, or an instruction and its reference. Do not treat the most visually prominent point as the only important content.
Use material whose intended use the responsible owner has resolved. A crop brief does not establish permission or turn an untraceable screenshot into evidence. Preserve a sample label or qualification that changes how the image should be understood.
Separate cropping from image sharpness
MDN's responsive-image guide distinguishes changing the composition for different layouts from supplying different resolutions of the same composition. These are separate decisions. A sharper copy of a crop still leaves out the content that was removed.
Record the proposed frame shape before requesting an asset. If the designer asks for a wide banner and a portrait card, inspect both compositions. “Make it responsive” is not enough information to decide what may disappear.
Keep a possible alternative open: show the complete image with surrounding space, change the frame, or prepare another permitted composition that preserves the same meaning. Do not force every source into every shape.
Work through four fictional crop decisions
Imagine Harbor Sample Studio has a comparison image showing two layout samples. The page asks the reader to compare them. Both samples are essential.
The fictional source is 1,200 pixels wide and 800 pixels high. Its essential content runs from horizontal position 100 to 1,100, and vertical position 200 to 600. Positions are measured from the top-left corner. This is an invented geometry exercise, not an inspected photograph or rendered browser result.
Scroll horizontally to compare all columns.
| Case | Proposed treatment | What the geometry establishes | Planning decision |
|---|---|---|---|
| C01 | Show the complete 1,200 × 800 source | The entire essential area remains inside the image | Candidate for a full-image placement; actual readability still needs checking |
| C02 | Center an 800 × 800 square crop | It retains horizontal positions 200–1,000, cutting part of the essential area on both sides | Reject this crop for the comparison task |
| C03 | Center a 600 × 800 portrait crop | It retains horizontal positions 300–900; moving this 600-pixel-wide crop cannot contain a 1,000-pixel-wide essential span | Change the frame or composition instead of guessing another focal point |
| C04 | Fit the complete source inside a 600 × 800 frame without cropping or distortion | At half scale, the image is 600 × 400; 400 vertical pixels remain outside it, or 200 above and below if centered | Candidate alternative; check whether the smaller details remain usable |
The calculations establish only these rectangular relationships. They do not prove that the original samples are legible, that the colors work or that a browser uses the planned treatment. All asset and browser checks remain NOT RUN.
C03 is the useful failure case. No horizontal adjustment can squeeze that entire essential span into the proposed crop. A better answer may be a different layout rather than another attempt to center it.
Tell the implementer which behavior you intend
MDN's object-fit reference describes the difference between preserving the whole image within a box and filling the box while clipping excess content. Its fill behavior can stretch content when the proportions differ. Those distinctions help explain the brief; they do not identify a feature in your builder.
Write the requirement in ordinary language first: “Keep both sample layouts visible together, without stretching them.” Then ask the implementer which actual mechanism will satisfy it in the current page.
MDN's object-position reference describes positioning an image's content within its box. That can help retain a particular area when a crop is otherwise suitable. It does not solve C03's geometry or establish that a background image uses the same controls.
Avoid naming a specific percentage or breakpoint until it has a reason in the actual design. The worksheet records the intended result and the implementation question separately.
Preserve the meaning across variants
If a different narrow composition is proposed, compare it with the original task. A close-up of one sample might be useful for a separate detail view, but it cannot silently replace an image whose job is comparing two samples.
Record which variant belongs to which placement and the source it came from. Keep the uncropped source reference so a reviewer can inspect what changed. A replacement image should not imply a demonstration or result that never occurred.
Review surrounding text at the same time. A caption saying “Compare both layouts” becomes inaccurate if the selected variant shows one. If the image is also a link, its destination needs to remain understandable.
Use the image-alternative guide for that separate text-equivalent decision. Alternative text does not restore missing visual comparison for a sighted reader, and a successful crop does not establish the correct alternative text.
Check the actual selected image in its page context
Give the reviewer the source, proposed variants, essential-area description and intended placements. On the implemented page, record the exact page version, viewport and image actually selected. A source file in an asset folder is not evidence that the page displays it.
Inspect wide and narrow placements and the transitions the implementation uses. Check the complete essential content, distortion, readable detail and any overlay that covers the image. If the actual frame differs from the plan, record its dimensions before deciding whether the source needs another crop.
Do not close a narrow-layout issue with a wider screenshot. Keep the original failing condition and check the changed version there. Separate image selection, visual meaning and other page behavior; this is not a full accessibility or performance evaluation.
Use the image-crop worksheet for C01–C04 and a blank asset/placement record. Bring it to the ClickFunnels AI workflow at /agents, subject to current access requirements. It grants no builder admission, image-generation permission, automatic crop behavior or publication. Resolve the first frame that cannot preserve the image's actual job.