Homepage Hero Options
Live hero: composition correction
Adam, after seeing the deployed hero at full size: the composition reads as two halves stuck together. Three genuinely different compositions below, using only the two supplied Pexels assets. Production has not changed and the current hero is still live. Desktop shots are full width here so composition problems are not hidden by scaling; each option has a clickable full-size prototype.
c374e46, serving on the apex and on www. The other compositions below are kept as
exploration and history. The second supplied Pexels asset is live too, further down the homepage in
the entity block beside the paragraph about working from what a family already has.Option 1: SHIPPED
Live homepage, after the spacing pass.
What changed from the prototype
A spacing pass rather than a straight copy. The hero was closing about 110px below the Amanda quote, which reads as unused page height rather than breathing room, so the section's bottom padding and the copy column's top padding were cut back and the composition now ends on the quote. Mobile order is headline, proposition, photograph, CTA, reassurance, quote, with a modest image height so the CTA stays close to the proposition.
Second Pexels asset
No new section was invented for it. The entity block's left column ended after one sentence and sat as a column of air beside three paragraphs; the photograph now fills it, directly beside the paragraph about the recordings, photographs, letters and half-finished journals a family already has. Hero is conversation, this is pages and material, the sample passage further down is the written outcome.
Current live hero (control)
Pexels asset 1, pexels-olly-3768131, cropped.
What changed compositionally
Nothing. This is what is deployed.
Where the Amanda quote sits
Below the photograph, extending the right column further.
Strength and weakness
Strength: the photograph is warm and the proposition is clear. Weakness: a hard vertical edge runs the full height at roughly the halfway point, the photograph is neither contained nor full-bleed, the left half carries a large dead zone under the price line, and the quote adds another block to an already heavy right side.
Option 1: Conversation, editorial
Pexels asset 1, pexels-olly-3768131, cropped.
What changed compositionally
The cream is the page and the photograph is an element on it. The image is contained with margin on every side, dropped well below the headline baseline so it shares no edge with the header, and the copy column is widened. The hard full-height edge is gone.
Where the Amanda quote sits
In the left column, under the price line, as a small italic annotation.
Strength and weakness
Strength: calm and unmistakably editorial, and the spaciousness now reads as deliberate because the image is framed rather than bleeding. Weakness: the image is smaller, so it carries less emotional weight, and the lower left still has air under the quote.
Option 2: Words and material
Pexels asset 2, pexels-cottonbro-6939474, recropped to keep both women AND the open pages.
What changed compositionally
A magazine opener. No vertical split at all: headline left and proposition plus CTA right across the top, then one wide letterbox photograph spanning the full content width beneath them.
Where the Amanda quote sits
Directly under the photograph as its caption, running the width.
Strength and weakness
Strength: the only option where the written outcome is visually present, and the wide band uses the full width instead of dividing it. Weakness: the letterbox crop clips the tops of both heads, and two people looking down at pages reads more as reading together than as somebody telling their story.
Option 3: Best hybrid
Pexels asset 1, pexels-olly-3768131, recropped taller for a narrow column.
What changed compositionally
The composition I would choose designing fresh. A tinted band holds the whole first viewport so it reads as one composition rather than two halves. The photograph becomes a tall narrow portrait column set at about the two-thirds line rather than half the page, and the copy takes the width it needs.
Where the Amanda quote sits
In the left column under the price line, inside the band, so it belongs to the composition.
Strength and weakness
Strength: the tint does the work the split was failing to do, and nothing has a hard edge against anything else. Weakness: a tinted hero band is a stronger commitment than the current flat cream, and the narrow column crops the conversation tighter than the wide version did.
What I changed in B, C and D besides the picture. The proposition was trying to explain three output formats before it had explained the service, so in every option except the control it reads: Amanda sits with someone and talks about their life, listens properly, and turns those conversations into a story that still sounds like them. The formats belong further down. The headline is untouched. The Amanda quote is kept in all three but demoted, as a caption in B and C and as a quiet line under the CTA in D. The sample passage is not deleted, it is removed from the hero; the proposal is that it moves down the homepage into the writing-proof area where it can be large and properly typeset.
A: Current live
Control. Unchanged.
What it does
Two large text blocks competing: headline, proposition, Amanda quote, two CTAs and a price line on the left; a full sample passage plus its disclosure on the right. Typographic rather than human, and the sample passage carries almost the same visual weight as the proposition.
What changed from A
Nothing. This is what is live now.
B: Conversation led
New temporary image, full-bleed right.
What it does
Leads with the conversation itself, which is what the service actually is. The older man is mid-sentence; the listener is seen from behind with no face visible, so nobody is implied to be Amanda or a client. Headline, proposition and CTA get the whole left column with room to breathe. The Amanda quote drops to a small caption under the photograph.
What changed from A
Proposition rewritten to explain the service before listing output formats. Sample passage removed from the hero. Quote demoted to a caption.
C: Material and people
New temporary image, full-bleed right.
What it does
Same structure as B but a different emotional emphasis: hands over an open album rather than a face. Warmer and more tactile, less about the interview and more about what a family already has. No faces at all, so it is the safest of the people-led options.
What changed from A
As B, with a different image.
D: Material led, split
Uses the approved photographs-on-a-table asset already on the site.
What it does
A true half-and-half split: photographs left, copy right. The most editorial of the four and the only one that does not need a new asset. Leads with the material rather than with people, so it is calmer and less specific about who is in the room. The Amanda quote sits quietly under the CTA, which gives it a real home without a caption.
What changed from A
As B. Image is one already live elsewhere on the site, so no new dependency.
Shipped: previous live hero to new live hero
Deployed 20 September 2026, commit 5f822bd, serving on the apex and
on www. The photograph is option E, cropped. The copy is treatment E-B. Mobile is proposition first.
Previous live
Two competing blocks of copy: proposition left, sample passage right.
New live
Photograph right, one proposition left, Amanda quote as the caption. Mobile leads with the headline.
Final proposition
Nobody remembers a whole life in one afternoon. Amanda knows what to ask, leaves room for the things that come back a week later, and writes it all into a life story that still sounds like them.
Why E-B over E-A and E-C
Scored against the eight things the proposition has to convey. E-A covered six. E-B and E-C both covered seven. E-B's gap is existing material, which the entity block immediately below already covers. E-C's gap is that you do not have to remember it all at once, which is the objection that actually stops people and is not covered anywhere else above the fold. E-B also opens with a human sentence rather than a list.
Final hero photography test
Option B’s composition is now the baseline and is not being redesigned. The only variable below is the photograph. A, C and D remain above as reference; the rejected image bank stays at the bottom of the page. The third supplied Pexels photograph is deliberately not here: the man reads too young for the intended parent relationship and the albums dominate the activity, so it is held as a possible inner-page image only and is not on the site.
B: Generated conversation
Generated image. The existing baseline.
The photograph
An older man mid-sentence, listener seen from behind with no face. Nobody identifiable and nothing claimed.
Strength and risk
Strength: the composition is proven and the man is unmistakably telling somebody something. Risk: the listener is anonymous and obscured, so it can read as an interview or a session rather than an adult child listening to a parent.
E: Pexels conversation
Supplied Pexels photograph, cropped.
The photograph
The older woman is visibly talking and the middle-aged woman is turned towards her and engaged. Cropped to remove the lower band, which takes the hand-holding almost entirely out of frame.
Strength and risk
Strength: the clearest read of the three, an older parent telling a story and an adult child listening, with a plausible age relationship. Risk: the arm around the shoulder is still visible and can tip towards care rather than conversation; it is outdoors, which is warm but less like a recorded session at home.
F: Pexels people and material
Supplied Pexels photograph, cropped hard towards the people.
The photograph
Cropped from the left and bottom so the two women dominate and the open book is reduced to a hint at the lower edge.
Strength and risk
Strength: warm, domestic, indoors, and the material is present without being the subject. Risk: both women are looking down at the page rather than at each other, so it reads as reading together rather than conversation, and the age gap is narrower than the intended parent and adult child.
E: the two mobile orders
Same photograph, same copy. Only the order changes.
Mobile 1, image first
Header, image, Amanda quote, then the headline. The photograph gets the whole first screen and does the emotional work, but the visitor meets a picture before they meet the proposition, and the quote sits between the two, which pushes the headline further down still.
Mobile 2, proposition first
Header, headline, proposition, then image, then CTA and quote. The visitor knows what Capture and Keep is before the photograph arrives. The trade-off is that the CTA now sits below the image and is off the first screen, and the photograph has less impact at a smaller height.
All six previous assets were recovered from
plan/retired-images/, control-panel/captureandkeep/ and the old Kept Words
mockup, and opened individually. None of the six is usable as a hero. That is not
because they are AI generated, which is fine as temporary imagery, but because of what they depict.
The two new candidates at the bottom were generated for this exploration and are not on the site.
 |
session.jpgPeopleREJECT |
Previously the session photo on the old Kept Words homepage mockup. Compositionally the strongest thing in the bank, and unusable. The woman is Amanda’s own likeness, generated from her reference photographs: same face, same style of patterned blouse as her real portrait. Putting it on the site says “here is Amanda with a client”, which is documentary evidence that the service has produced something. A real photograph of Amanda already exists. Real shot that replaces it: Shot 4: Amanda mid-session, from behind or side-on so the other person is not identifiable. |
 |
hands.jpgPeople and materialREJECT for hero, revisit later |
Previously on the homepage and /life-story. The most beautiful image in the bank and the biggest temptation. Two problems: the book is foil blocked KEPT WORDS, a brand that no longer exists, and it shows a finished hardback resting in an elderly person’s hands, which reads as a delivered client project. That is the single claim this site most needs to earn rather than borrow. The branding could be edited out; the implication could not. Real shot that replaces it: Shots 1 to 3: the real finished book, closed, open, and a close-up of the binding. |
 |
book.jpgMaterialREJECT |
Previously the homepage book visual. Carries an invented customer name in gold on the cover, “Ronald James Fielding”, plus the retired KEPT WORDS mark. Inventing a named person on an artefact is the clearest breach in the bank and is why this one was deleted rather than relabelled. Real shot that replaces it: Shot 1: a real finished book, front cover, in daylight. |
 |
spread.jpgMaterialREJECT |
Previously the book-spread visual. The body text is AI gibberish and does not resolve into words at any size, and the caption names the same invented family, “the Fielding family, summer 1955”. Two separate fabrications in one image. Real shot that replaces it: Shot 2: the same book open, a photograph on one side and text on the other. The most persuasive image the site could have. |
 |
voucher.jpgMaterialREJECT |
Previously on /gift-voucher. KEPT WORDS on the book, lorem gibberish on the card, and it depicts a keepsake pack that does not exist. The site stopped describing a keepsake box on 20 Sep because nobody had decided it, so this image now contradicts the live copy as well. Real shot that replaces it: Shot 6, and only if a real voucher object ever exists. |
 |
amanda-ai-standin.jpgPeopleREJECT |
The original AI stand-in for Amanda’s portrait, replaced by her real photograph on 7 Sep. Superseded. There is a real photograph of Amanda and it is already on the site. Never substitute a generated likeness for a real person who has supplied one. Real shot that replaces it: Already replaced. |
 |
cand-conversation.jpgPeopleNEW, used in option B |
Generated for this exploration. Not on the site. Depicts the subject matter rather than the service: an older man talking, a listener with no face visible. Nobody is identifiable, nobody is implied to be Amanda or a customer, and no finished product appears. No text anywhere in frame. Real shot that replaces it: Shot 4, or a real family conversation photographed with permission. |
 |
cand-album.jpgPeople and materialNEW, used in option C |
Generated for this exploration. Not on the site. Hands only, no faces at all, over an open album. The album photographs are small and indistinct and there is no legible writing anywhere, which was the defect that got an earlier letters image regenerated. Real shot that replaces it: Shot 7: real family material on a table, staged with permission. |