Should I vibe code
Visual portfolio websites for designers, illustrators, and photographers
Nobody notices your layout. A photographer notices the second your resizer throws away the colour profile.
?
Their verdict, the Professional price and the build-time estimate come from their entry, MIT-licensed. Checked 2026-08-04.
?
Our verdict, the regret score and everything below it. Editorial and unsponsored — nobody can pay to be moved.
The honest answer
why the verdict is what it is
Your own portfolio is a Saturday, and a good one. Static generator, a grid, a lightbox, a domain — the agent will hand you something better looking than most templates before lunch. Carbonmade is a different object, and the difference is entirely about pictures. It is unlimited 4K uploads from people whose files come out of Lightroom at forty megabytes, turned into responsive derivatives that load on a phone in a café while an art director decides whether to email you. That means a resize pipeline that preserves colour management, strips the GPS coordinates out of the EXIF without stripping the copyright field, generates AVIF and WebP with sensible fallbacks, and never once hands back a JPEG that looks flat. It also means bandwidth someone pays for, forever, on a site that is the front door to your income. None of that is dangerous. All of it is the part you would still be building in November.
What actually breaks
not "if". the specific failures.
- Colour. A camera JPEG is often Adobe RGB or Display P3 with an embedded ICC profile, and most resize pipelines drop the profile while keeping the numbers — which renders every image visibly desaturated. A photographer spots this instantly and cannot tell you why in technical terms, only that the site looks wrong
- EXIF. Strip nothing and you publish the GPS coordinates of the room the shot was taken in, which for a portrait photographer working from home is their address. Strip everything and you have removed the copyright and creator fields from their work
- Bandwidth, which is invisible until it isn't. Full-bleed hero images and autoplaying showreels are the house style in this category, and one project going around design Twitter is a bill or a throttle, whichever your host chose
- The derivative pipeline: srcset breakpoints, AVIF and WebP with fallbacks, orientation from the EXIF rotation flag that half the tooling ignores, and regenerating all of it the day you change a layout
- Video. The moment an animator signs up, you are transcoding, storing and streaming, and the storage cost stops being rounding error
- Uptime you had not thought of as uptime. A portfolio is checked once, by someone with a shortlist, at a moment you will never know about
- Originals, if the uploader became the archive. People do delete the local copy after they have "put it online", and your object store is now the only place a 2021 campaign exists
- The editor, if anyone but you uses it. Carbonmade's whole pitch is launching in minutes without touching code, and that is a UI problem measured in months
Is that you?
the verdict is a default, not a law
- It is your own portfolio, built as a static site, deployed to a host with a free tier and a CDN in front of it
- The originals live in your normal photo library and the site is a published copy, so the site is never the only copy of anything
- You are comfortable regenerating the whole thing from source whenever you change the layout, because that is the actual maintenance model
- There is no upload form, no login, and no second person who needs to change anything without you
- You are hosting other creatives' portfolios, which turns a hobby into an obligation to be up during somebody else's job hunt
- The site is where the high-resolution originals live and no backup exists elsewhere
- You have not decided, explicitly, what your pipeline does with EXIF — publishing a home address by accident is a one-line default
- The client needs to swap a project on a Sunday, because at that point you are not building a portfolio, you are building an editor
If you build it anyway
the checklist, then the prompt that enforces it
- Convert to sRGB explicitly on resize rather than dropping the ICC profile. Test with an Adobe RGB source and compare the output side by side before you believe the pipeline works.
- Strip GPS and camera serial from EXIF, keep copyright, creator and description. Make that policy a single function with a test, not a flag you passed once.
- Honour the EXIF orientation flag and bake the rotation into the derivative. Sideways portraits are the most common bug in every homemade gallery.
- Generate derivatives at build time, not on request. A resize-on-the-fly endpoint with a URL-controlled width is a free CPU-exhaustion button for anyone who finds it.
- Put a CDN in front and set long cache headers with content-hashed filenames, because the traffic pattern here is a long flat line with one enormous spike in it.
- Never let the site be the only copy. Uploads land in object storage with versioning on, and originals are backed up somewhere the deploy pipeline cannot touch.
- Keep the content in plain files — Markdown or JSON plus images in a folder — so the site is portable to any generator and the exit is a copy command.
I am building a portfolio site for creative work: galleries of large images,
some video, a custom domain. The interesting work is the image pipeline, and
the way it fails is that the pictures come out subtly wrong or leak something
they shouldn't. Do it in this order.
1. Content lives in plain files — Markdown or JSON plus an images folder — not
in a database. I want the exit to be a copy command.
2. Build the image pipeline before any layout. For each source file produce a
fixed set of widths plus AVIF and WebP with a JPEG fallback, at build time.
No resize-on-request endpoint that takes a width from the URL.
3. Convert explicitly to sRGB and embed the profile. Do not simply drop the
input ICC profile — show me a before/after with an Adobe RGB source so I can
see there is no colour shift.
4. Bake EXIF orientation into the output. Sideways portraits are the failure I
expect from a first version.
5. Strip GPS, camera serial and lens serial from metadata; keep copyright,
creator and description. One function, with a test, and tell me in the
README that this is why.
6. Emit correct srcset and sizes, set width and height on every image so
nothing reflows, and lazy-load everything below the fold.
7. Video is transcoded ahead of time to a capped bitrate and never autoplays
with audio. Tell me the storage and egress implications before you add it.
8. Content-hash every asset filename and set immutable long cache headers.
Assume one project goes viral once and the rest of the year is quiet.
9. The build must be reproducible from the source folder alone. If I delete the
deployed site, one command puts it back.
10. Do not build an upload form, a login, or a CMS. If I ask for one, remind me
that I am then building an editor for somebody who does not write code, and
that Carbonmade charges $14.99 a month for exactly that.That one keeps you out of trouble. For the prompt that actually builds it, canivibecodeit.com has one.
their build prompt ↗Or don’t build it
the boring option, and the way back out
If the site belongs to someone who will need to change it without you, or if the person whose work it is bills by the hour and does not want to spend Saturday reading about ICC profiles. Fifteen dollars a month is hosting, bandwidth, a CDN, a resize pipeline that photographers have already complained at, and — the real product — a drag-and-drop editor that a working illustrator can use to swap a project at 11pm before an interview. Your version can be prettier. It cannot be that.
$14.99/mo is cheaper than your weekend.
Excellent, and it is the reason this category is a comfortable DEMO ONLY. Keep the content in flat files and images in a folder and the whole site is portable to any static host in an afternoon; the domain is yours, the DNS is yours, and there is no export step because there was never an import. The only exit that hurts is the one where uploads went into a bespoke schema and the originals were deleted from the laptop afterwards — so decide on day one that the repository is the archive and the deployed site is a build artefact.
Open-source visual website builder, if what you actually want is the editor rather than the site.
Embeddable drag-and-drop page-builder framework, worth reading before writing your own canvas.
Questions
Squarespace and Tilda are already here. What is different about this one?
Squarespace's entry is about the editor a non-technical client uses on a Sunday, and Tilda's is about five hundred blocks that already behave on a 320-pixel phone. Carbonmade's distinguishing feature is the payload: a portfolio is four hundred very large photographs, and everything that makes it hard is downstream of that. Different bottleneck, same band.
Is the colour profile thing really worth a whole bullet?
For this audience, yes. Cameras and editing software routinely output Adobe RGB or Display P3, and a naive resize that drops the embedded profile while keeping the pixel values makes every image render noticeably flat. The person whose portfolio it is will see it in two seconds and conclude your site is broken, because from their point of view it is.
What about the contact form?
It is the one piece of server you cannot avoid, and it will be found by spam within a week of the domain going live. Use a hosted form endpoint or your host's function with a rate limit and a honeypot, send to a real inbox, and check that it works every few months — a silently broken contact form on a portfolio is the most expensive bug in this entry and the only one that costs you actual work.
Every week, someone ships something they shouldn’t have.
New verdicts, the worst thing that landed in the trap, and the occasional incident report. No other email, ever.
Cargo's $19 is a foundry licence with a website attached. Copying the woff2 into /public is not the same deal.
Your portfolio is built to be crawled. The gallery beside it holds a campaign that has not launched yet.
Nobody wants a website builder. They want to change the opening hours on a Sunday without texting you.
last reviewed 2026-08-04 · verdict is editorial and unsponsored · shared entry data from canivibecodeit under MIT · not legal advice