Video walkthroughs · Claude Code Club

How To Make A Website Look Expensive

13 minute read · Companion guide to “Claude Fable 5 builds INSANE $10,000 Websites! (with 1 prompt)”

Expensive is not a budget. It is a small set of craft decisions, made in a specific order, and each one narrows the ones after it. That is why a site can cost five figures and look almost plain, and why a site with three animations, four fonts and a stock photo of a smiling team reads as cheap no matter how much work went into it. Once you can name the six decisions, you can make them deliberately instead of hoping the result feels premium.

Watch the full build: a luxury car site assembled piece by piece, then a second site from one image

Expensive is a look you can specify

Sites that look costly share a quality that is easy to feel and hard to name: everything on the page agrees with everything else. The color agrees with the product. The motion agrees with the message. The photography looks shot on the same afternoon in the same light. One idea, restated in different materials.

Cheap sites break that agreement in small ways. The hero says calm luxury and the button is emergency red. Two images were generated an hour apart and one has warm evening light while the other is lit like a showroom. Nobody consciously notices any single break. Everyone feels the sum of them.

So the whole craft is a sequence that keeps agreement from breaking. Six pieces, in this order, because each one inherits from the one before:

1. The core messaging. The one thing the site says.
2. The scroll style. How the page moves, decided before any copy exists.
3. The color palette. Pulled from a scene, not from a brand guideline.
4. The copy. Short headlines and specific proof, written to the message.
5. The photo and video. Five assets, all lit the same way.
6. The spine. One final pass that makes every piece agree with the others.

Run them out of order and you feel it immediately. Copy written before the scroll style is decided gets rewritten, because a page that plays video as you scroll needs three word captions and a static hero needs sentences. Assets generated before the palette exists come back in the wrong light and get thrown away.

Piece 1: Say one thing, and let it govern the other five

A premium site says one thing. That is the entire discipline. Saying one thing reduces overwhelm, and it speaks directly to the person who is actually going to buy instead of hedging toward everyone who might.

The fastest way to get there is to answer four questions before anything is built, and to answer them in one line each. Long answers are a sign the message is not decided yet.

• What is the product, in one line? Not a description. A name and a category.
• Who is it for, and what should they feel when they land? The feeling matters more than the demographic, because the feeling is what the motion and the palette will be built to produce.
• What is the one thing they should remember after they leave? This is the message. Everything else on the page serves it.
• Are you generating the photography and video, or using stock? Answer this now, because it changes the shape of the build rather than being a detail you sort out later.

In the walkthrough the product is an Audi R8 and the answers took about forty seconds. Premium buyers and thrill seekers. They should feel excitement and speed. The one thing to remember: this is your dream car and it could be in your driveway this week. Notice that the memorable line is not a feature. It is a moment the buyer can picture themselves inside.

The first build off those answers is deliberately ugly. Plain type, no color, no images, one headline. That is correct, and you should not fix it yet. It exists to prove the message works with nothing propping it up. If the line does not land as black text on white, no animation will save it, and you found that out for free.

Piece 2: Choose the scroll style before you write a word

How the header moves is the first thing anyone sees, and it sets how they feel about the product before they have read a single word. It is also the decision most people leave until last, treating motion as polish applied over finished content. That is backwards. Movement is half of what makes a site feel expensive, so it is a structural decision, not a finishing one.

There are four styles worth knowing, and the product tells you which one it wants:

• Looping video. Best when the product is already in motion or the feeling is atmosphere rather than detail. Cars, drinks, travel, anything where mood sells.
• Scroll to play. The video advances as the visitor scrolls, so they control the reveal. Best when the product has a transformation to show, and best of all when the copy can track the frames.
• Exploded view. Components separate and float apart as you scroll. Best for engineered objects where the value is inside: watches, headphones, tools, anything where the buyer wants to see the build quality.
• Static hero with a slow move. One strong image with a light sweep or a slow push. Best when the product is a person, a service, or anything with no visual drama to mine. Easy to underrate, and it never looks cheap.

The test that stops you from picking on vibes: write one sentence defending the choice against the message from piece one. If you cannot defend it in a sentence, you picked the one you thought looked coolest, which is how sites end up with an exploded view of something that does not come apart.

This is also where page structure gets set, because motion determines structure. Hero, thesis, feature blocks, gallery, closing ask. Decide the scroll style and that skeleton falls out of it, which is why nothing is written yet.

Piece 3: Let a scene choose the palette, not a brand guideline

This is the piece that generalizes furthest past websites, and it is the one worth stealing even if you never build another page. The normal way to pick colors is to describe the brand. Sporty. Premium. Trustworthy. Approachable. Those words produce the same palette every time, which is why so many sites look like each other.

Do this instead. Write one sentence describing who is using the product, where, and in what light. Then take the palette out of that scene.

The sports car site got this sentence: a buyer takes his new R8 out alone at dusk on an empty canyon road just after sunset. Everything follows from it. Dusk means the page is dark. Low sun means warm highlights against cool shadow. An empty road means space and restraint rather than density. None of that was chosen as a color. It was chosen as a moment, and the colors were what the moment already contained.

Another example, for a completely different product: a smith pulling steel from the fire after dark. Same mechanism, and you can already see the page. Near black background, orange heat as the only saturated color, metal grays. A brand guideline would have said industrial and rugged and produced something forgettable.

The reason this works is that a scene carries a light source, and a light source carries a palette that is already internally consistent because real light is. Adjectives do not carry light. They carry taste, and taste averages out to the middle.

The fork worth naming: dark pages read as premium and cinematic, and they hide flaws, but they punish long copy and dense information. Light pages read as clean and confident and can carry more text, but they expose every misalignment and every asset that was lit slightly differently. If the site is short and mood led, take the dark. If it needs to explain anything, take the light and accept that you will have to be more precise.

Piece 4: Write copy as proof, not as adjectives

Expensive sites are quiet. That is the whole voice, and it is the opposite of what most people write when they are trying to sound premium. The rules are narrow enough to be checkable:

• Headlines stay short. If it wraps to three lines on a phone, it is not a headline yet.
• Captions run three to five words, with maybe one accent word carrying the emphasis.
• Claims get specific to the point of discomfort. Two thirds less acid beats smooth and delicious every single time, because a number can be checked and an adjective cannot.
• No section restates the headline above it. Restating is what filler does when it has nothing to add.
• No dashes standing in for thinking. If a sentence needs one, the sentence has two ideas and wants to be two sentences.

Here is the copy the car site produced. Production of the R8 ended in 2024, so every R8 that will ever exist already does, and one of them can be in your driveway by Friday. Then the engine: 8,700 RPM. Then a scene: drive it to the office on Monday, then take the long way home through the canyon and arrive grinning.

Look at what is absent. No stunning. No unparalleled. No engineered to perfection. Every line is either a fact or a moment, and both are things the reader can do something with. Adjectives ask the reader to take your word for it, and a reader who is about to spend real money will not.

Piece 5: Generate five assets, not fifty

Generated photography and video are what separate these sites from templated ones, and not because generation is impressive. It is because generated assets can be directed. You can put the camera exactly where the story needs it, in exactly the light your scene sentence established, which is something no stock library will ever give you.

The restraint is the craft. The whole car site runs on five assets: two videos and three images. Same lighting on all five, same palette, a mix of close detail and wide context. Five assets that agree beat twenty that do not, and the sites that look cheapest are almost always the ones with the most images.

A five asset set that works for almost any product:

1. Hero video. The establishing shot, in the light from your scene sentence. This is the one that earns the highest quality setting you can afford.
2. One wide image. The product in its environment, giving the eye somewhere to rest after the hero.
3. One macro detail. Texture, material, craftsmanship. This is the shot that says the thing is real.
4. A second short video. Motion or a single moving detail, placed at the point in the scroll where attention normally drops.
5. One closing image. Product or moment, matched to the closing line, sitting near the ask.

Connecting a render tool directly to your terminal is what makes this one pass rather than a tab-switching afternoon. Two paths. A hosted service like fal.ai is the easy start: make an account, create an API key, hand it over once, and the cost stays visible because you pay per credit used. The alternative is connecting a generation service over MCP, which gives you a wider set of models from inside the session. Take the API key path for your first build, because one failure mode is easier to debug than two.

Piece 6: The spine pass that makes everything snap together

Six good pieces do not automatically make a good page. Something has to go back over the whole thing and enforce agreement, and that final pass is the difference between a site that looks assembled and one that looks designed.

The spine checks one thing in five places: does this agree with the message from piece one?

• Section order. Does the sequence still build toward the one thing they should remember, or did a section get added that serves something else?
• Palette consistency. Every accent, border, button and text color traced back to the scene sentence, with nothing left over from the plain first draft.
• Asset placement. Every image and video sitting at the beat it was made for, sized to the attention it deserves.
• Copy rhythm. Short where it should be short, nothing restating anything above it.
• Motion consistency. Every transition moving at the same speed and with the same character. Mixed easing is one of the loudest cheap signals there is, and almost nobody can name it.

This is also the piece that turns a one-off build into a repeatable one. Because the spine is a fixed check rather than a creative act, it can run identically on any product, which means the whole six-piece sequence becomes a thing you run rather than a thing you invent. That repeatability is the actual reason several finished sites in one afternoon is possible, not raw model speed. The model was always fast. What was slow was deciding everything again from scratch each time.

How the same six pieces run from a single image

Once the sequence is fixed, the inputs collapse to almost nothing. A product photo and a paragraph of details is enough, because the six pieces know what to ask for. The second build in the walkthrough is a pair of Air Jordans, and the inputs were the image plus details copied straight off a product page. The interview got skipped, since the answers were already in what was handed over.

Two things change when you run it this way. The work parallelizes: three drafts get produced at once from the same six decisions, then scored against design criteria, and the best one survives. And the checking becomes part of the build rather than something you do afterward with tired eyes.

The output was a break-apart render at the top, photoreal close-ups with visible texture, a light sweep across the shoe, and finished copy. From an image and a paragraph. The reason it works is not that the request was clever. It is that the six decisions were already made and only the inputs changed, which is exactly what makes batching possible.

The failure modes that make a site read cheap

Each of these is a break in agreement, and each one is cheap to avoid and expensive to fix after the fact.

• Skipping the ugly first draft. It is the only moment you can tell whether the message works on its own. Once color and motion are on top, you cannot see past them.
• Picking the scroll style after the copy is written. Either the copy gets rewritten, or worse, the motion gets bent to fit copy written for a different kind of page.
• Choosing colors from adjectives. Sporty and premium produce the same palette everyone else has.
• A scene sentence with no light in it. Then dark or light gets decided by preference and the assets have nothing to match.
• More than five or six assets. Every extra one is another chance for the lighting to disagree with the rest of the page.
• Adjective copy. Stunning, unparalleled, engineered to perfection. All of it reads as a product that needs to argue for itself.
• Mixed motion character. Some transitions snappy, some floaty. Nobody names it, everybody feels it.
• Too many fonts. Two is plenty. Three is a decision that needs defending and usually cannot be.

Build three of these in one afternoon

The point of a fixed sequence is volume, and volume is what makes you good at it. Pick three products you can find a clean photo of, in deliberately different categories, so you feel the palette and motion decisions diverge.

1. Answer the four questions for product one, in one line each. Stop when the third answer is a single memorable sentence.
2. Build the plain, ugly first version. Read the headline as black text on white and decide honestly whether it lands.
3. Pick the scroll style and defend it in one sentence against the message. Let the page skeleton fall out of that choice.
4. Write the scene sentence. Who, where, in what light. Derive the palette and the dark or light decision from it, and apply both.
5. Write the copy to the rules: short headlines, three to five word captions, specific claims, no restating.
6. List your five assets with a purpose sentence for each, then generate all five in the light your scene sentence set.
7. Run the spine pass over the whole page: order, palette, placement, rhythm, motion character.
8. Repeat for products two and three. By the third one the decisions take minutes, because you are choosing from a known set instead of inventing.

Three finished sites in three categories is a portfolio, and it answers the only question anyone asks before hiring you: can you do it for something like mine. One site is a demo. Three is proof that you have a process, and a process is the thing that actually gets bought.

Everything in this guide is runnable by hand today, and running it by hand a few times is how you learn to see the six decisions in other people's sites, which is the skill that outlives any particular tool.

Read this online at claudecodeclub.ai