Four different page shapes, and only two of them put prompt text on the page. Measured on 13 October 2026 with one script over 144 pages belonging to the three faymas domains: a 100-page sample of faymas.in prompt pages carries a median 911 visible words and zero input fields; faymas.net's 23 model pages carry a median 710 words plus one textarea, and that textarea has no text in it on all 23; faymas.net's 11 gallery pages carry 438 thumbnail cards and no prompt text at all; faymas-in.com's three posts carry a median 3,677 words of copy-paste list. Not one of those 144 pages contains a single HTML table.
This guide analyses published prompt text. No generation tests were run.
How we measured it
One script, one set of definitions, three domains. That is the whole design, and it is the part that matters — comparing page weights between sites is meaningless if each site gets measured differently.
The script fetches a URL, strips script, style and noscript blocks, removes tags, and counts what is left as "visible words". It counts bytes on the raw response. It counts input, textarea, button, image and table elements in the served HTML. Then it writes down where the first four-digit year sits in the visible text, because a year in the last 15% of a page is a copyright line, not a date.
The samples: 100 prompt pages drawn from faymas.in's sitemap, the complete 41 URLs in faymas.net's sitemap, and all three posts that faymas-in.com's WordPress API returns. One hundred and forty-four pages. Every figure below comes out of that run and nothing else.
Four page shapes, side by side
Sort the 144 by what the page does and they fall into four groups. No group resembles another.
| Page type | Pages | Bytes (median) | Visible words (median) | Input field | Textarea | Images (median) | Tables |
|---|---|---|---|---|---|---|---|
| faymas.in prompt page | 100 | 563,412 | 911 | 0 of 100 | 0 of 100 | 14 | 0 |
| faymas.net model page | 23 | 213,726 | 710 | 14 of 23 | 23 of 23 | 4 | 0 |
| faymas.net gallery page | 11 | 345,021 | 834 | 11 of 11 | 0 of 11 | 56 | 0 |
| faymas-in.com post | 3 | 106,890 | 3,677 | 3 of 3 | 3 of 3 | 5 | 0 |
Read the input column first. It separates the sites more cleanly than any word count does. faymas.in prints a prompt and gives you nothing to type into — zero inputs and zero textareas across all 100 pages. faymas.net's model pages do the opposite: 23 of 23 ship a textarea. faymas-in.com's three posts carry both, and those are a comment form, not a tool.
Then read the image column. The gallery pages carry a median 56 images against 4 on the model pages. Those two numbers describe two jobs: one page shows you what came out, the other asks you what to put in.
The empty box on faymas.net
Every one of the 23 model pages ships a textarea. Every one of them ships it with nothing inside.
We checked the text content between the opening and closing tags on all 23. Zero had text. So the page that faymas.net builds around each of its models — the page that sits at /gpt-image/gpt-image-2, at /seedance/seedance-2-0, at twenty-one other model routes — contains no prompt until a visitor types one. It is a workspace, not a document.
That is a defensible product choice. It is also worth stating plainly, because the page looks like a prompt page and is indexed like one. What it holds is a form, a model name, a short description, and example titles. The prompt arrives later, from you.
faymas.in works the other way round. Its page is finished before you touch it: the prompt is printed, and the only interactive controls are buttons.
Weight: how many bytes sit behind one word
Divide bytes by visible words and the three sites land in three different places. This ratio is the fairest single number we have, because it punishes markup that delivers nothing.
| Page type | Bytes per visible word (median) | Range |
|---|---|---|
| faymas.in prompt page | 621 | 312.6 – 818.1 |
| faymas.net gallery page | 396 | 357.6 – 745.8 |
| faymas.net model page | 300 | 260.8 – 311.9 |
| faymas-in.com post | 29 | 27.8 – 110.4 |
faymas.in spends 621 bytes of HTML per word a reader can see. faymas.net's model page spends 300 — less than half — while also handing over a working input. The WordPress posts spend 29, roughly one twenty-first of faymas.in's figure, because a plain blog post is mostly text.
The spread inside each group is the other half of the story. The 23 model pages sit in a narrow band from 260.8 to 311.9 bytes per word. The 100 faymas.in pages spread from 312.6 to 818.1. A narrow band is what a fixed template produces; a wide one means the payload varies, which is what happens when each page carries a different amount of prompt text inside the same shell.
The year that is not a date
All 41 faymas.net pages carry a year somewhere in their visible text. That sounds like the freshest site of the three. It isn't.
We recorded where the first year appears. On 34 of the 41, it sits in the last 15% of the page text — the footer, reading as a copyright line. Seven put a year in the body. So faymas.net looks dated everywhere and states a date almost nowhere.
faymas.in inverts it. Only 33 of its 100 sampled pages carry a year at all, but 32 of those 33 carry it in the body, attached to a published date. Fewer pages, real dates.
Neither site is lying. A copyright line is not a claim about freshness. But if you are trying to tell whether a prompt is current, the count of years on a page is the wrong number to look at, and on faymas.net it is the only one on offer.
Zero tables in 144 pages
This was the finding we did not set out to look for. Across 100 faymas.in prompt pages, 41 faymas.net pages and three faymas-in.com posts, the number of HTML tables is zero. Not "few". Zero, on every single page.
We checked the count again at the end of the run, in case the parser was at fault, by grepping the raw responses for the opening tag. Still zero.
So none of the three sites tabulates anything: no model comparison, no pricing matrix on the pages we measured, no parameter reference, no side-by-side of output settings. Everything is prose, cards, or a form. For a subject built entirely of parameters — aspect ratios, model versions, credit costs, reference-image requirements — that is a strange omission, and it is the same omission on all three.
The gallery pages come closest to structured data and still do not get there. They carry 438 cards across 11 pages, and 111 of those cards are flagged as requiring a reference image. That flag is a fact about each prompt that would sit perfectly in a column. It sits in a card label instead.
What this measurement cannot tell you
We read server responses. We did not run a browser. If faymas.net fills its textarea after the page loads, we would not see it, and our claim is limited to what the server sends — which is also what a crawler that does not execute JavaScript receives.
We did not create accounts on any site and did not generate anything. No credit was spent, no output was inspected. Everything here describes the page as delivered, not the product as experienced.
The faymas.in figure is a 100-page sample of a catalogue that numbered 16,762 prompt URLs when we last counted it on 12 October 2026. A sample of 100 cannot speak for 16,662 others. The faymas.net and faymas-in.com figures are complete: that is every URL either site publishes.
Page counts change. faymas.in's catalogue grows daily, and faymas.net's sitemap held 41 URLs on both 4 October and 13 October 2026, which is as much stability as we can claim.
The gap the numbers point at
Set the four shapes next to each other and the missing shape is obvious. One site prints prompts and cannot take input. One takes input and prints nothing. One prints long lists and has no structure. One shows 438 thumbnails and no text.
What none of them does is put the same facts in the same columns twice. Model name, credit cost, whether a reference image is needed, what the output looked like, when it was published. All four exist on these sites. None of them are ever lined up.
That is the part worth noting for anyone writing in this space, including us. The supply is not thin — 144 pages measured here, and one of these sites indexes a catalogue in five figures. What is thin is the layer where someone writes down the parameters and lets you compare them. Prompts, it turns out, are the easy half.
FAQ
Do the three faymas sites build the same prompt page?
No. Measured on 13 October 2026, faymas.in prompt pages carry a median 911 visible words and no input fields, faymas.net model pages carry a median 710 words plus a textarea that is empty, faymas.net gallery pages carry thumbnail cards and no prompt text, and faymas-in.com posts carry a median 3,677 words.
Does a faymas.net model page contain a prompt?
No prompt text. All 23 model pages in its sitemap ship a textarea, and none of the 23 has any text inside it, so the page holds a form until you type into it.
Which faymas site has the heaviest pages?
faymas.in, at a median 563,412 bytes and 621 bytes per visible word. faymas.net's model pages are lighter at 213,726 bytes and 300 bytes per word, and the three faymas-in.com posts are lightest at 29 bytes per word.
Do these pages show a publication date?
Rarely in the body. All 41 faymas.net pages carry a year, but on 34 of them it appears only in the footer as a copyright line. On faymas.in, 33 of 100 sampled pages carry a year and 32 of those carry it in the body.
How many pages were measured, and how?
144 pages: 100 sampled faymas.in prompt pages, all 41 URLs in faymas.net's sitemap, and all 3 faymas-in.com posts. One script stripped script and style blocks, removed tags, counted the remaining words, and counted elements in the served HTML.
Can I reproduce this?
Yes, three commands. Fetch each site's sitemap and count the URLs; download a page with curl; strip tags and count words, then grep for the opening table tag. The last one returns zero on all three sites.