Novus Examples: free sample and dummy files for every format

Today we are adding Novus Examples to the Novus Stream Solutions portfolio. It is a free, browser-first library of sample and dummy files in every type and format we could reasonably cover, and the idea behind it is almost embarrassingly simple: when you need a test file, you should be able to download a real one in a couple of seconds, without hunting through old folders or signing up for anything. Think of it as a Lorem Ipsum for files, except that instead of placeholder text you get a genuine PNG, a genuine PDF, a genuine MP4, and, when you need one, a genuine zero-byte or deliberately corrupt file to make sure your software fails gracefully.

Almost everyone who works with software has hit this wall. You are building an upload form and need to confirm it accepts a JPG but rejects a 200 MB video. You are testing a converter and want to see what happens to a transparent PNG on the way to WebP. You are checking a background remover against a hair-heavy portrait, or a parser against a malformed JSON file. In each case the work is blocked on the same small thing: a suitable sample file, in exactly the right format, that you do not currently have to hand. Novus Examples exists to remove that particular friction, so the file you need is a download away rather than a detour.

Test files tend to live nowhere. They accumulate on desktops with names like final2.png and untitled.mp4, they get pulled from random corners of the web with no idea what is actually inside them, and they go missing the moment you switch machines. The result is a quiet, recurring tax on anyone who builds or tests software: every time you need a specific kind of file, you improvise one, or you export something from another tool, or you settle for the wrong sample because finding the right one is more trouble than the test is worth.

That improvisation has real costs. A test is only as trustworthy as the input you feed it, and a sloppy sample produces a sloppy result. If you check an image pipeline only against ordinary sRGB photos, you will not notice the day it mishandles CMYK or chokes on a grayscale scan. If you never try a zero-byte upload, you will not learn how your form behaves until a real user sends one. Good testing depends on having the awkward cases on hand, and the awkward cases are exactly the ones nobody keeps a copy of.

Novus Examples treats the sample file as something worth organizing properly rather than something to scrounge. Files are grouped by type and format, labeled by what they are meant to exercise, and available to download one at a time in the browser. The goal is that the right file for the test you are running is easy to find on purpose, so you spend your attention on the software you are checking rather than on the errand of assembling an input for it.

The library spans the formats people actually work with day to day, and it deliberately reaches past the common ones into the formats that break things. It has grown well past the handful of buckets it opened with on launch day: as of the July 31, 2026 release gate it holds 2,558 files organized into seventeen categories across 221 distinct extensions, so you can go straight to the kind of file you need.

One thing sets these samples apart from the test files you would otherwise scrounge from around the web: they are generated, not collected. Every file is produced deterministically to an exact specification rather than pulled from somewhere with unknown provenance, which makes the catalog copyright-free by construction. There is no stray photo or borrowed document with a license to worry about; a sample you download is a file made to be a sample, and nothing more.

Because the files are generated to spec, each one ships with a spec or property sheet describing exactly what it is: the noise levels baked into an image, the timestamps and encodings in a document, the sample rate of an audio clip. You are never guessing what is inside a test file, which is the usual failing of a scrounged sample. When a tool behaves oddly on an input, the property sheet tells you precisely what that input was, so the result reads as a signal about the tool rather than a mystery about the file.

Many samples also come as paired fixtures, two files meant to be used together. A clean version next to a deliberately messy one, a colour image next to its greyscale twin, a clean source document next to a scanned copy of it. Pairs are what let you test the transformation between two known states rather than a single file in isolation, which is exactly what you want when you are checking OCR, colourization, visual diffing, or any process that turns one version of a file into another.

Novus Examples began as a place to browse a category and download a file, and that is still the core. But it has grown a set of surfaces around that core for the work people actually do with sample files. There is a Search for going straight to a file by name or property instead of clicking through categories, and there is purpose-based navigation that lets you browse by what you are trying to do rather than by format: OCR testing, CSV parsing, form testing, colourization, ASR, visual diff and regression, schema and OpenAPI validation, JWT and JWKS handling, and more. You pick the job and the library shows you the fixtures built for it.

Two of the newer surfaces go beyond handing you a static file. An Editor lets you adjust a sample in the browser before you download it, and a Visual Template Studio lets you create documents locally from the template library rather than only taking what is on the shelf. Both keep the work client-side, in keeping with the rest of the tool. There are also Live Test Targets: real-world testing fixtures you can point a tool at rather than only files you download, which is what you want when the thing under test needs something to hit rather than something to open.

Between the browse-and-download core, the Editor and Visual Template Studio, Search, purpose-based browsing, and the Live Test Targets, the library has quietly become a workbench for test inputs rather than just a shelf of them. You can still open the site, find a format, and grab a file in a couple of seconds; the newer surfaces are there for the moments when a plain download is not quite the shape of what you need.

Novus Examples pairs most naturally with Novus Convert, our file converter at convert.novusstreamsolutions.com. A converter is one of those tools you cannot really evaluate without inputs, and the honest way to evaluate it is to feed it the formats you care about and inspect what comes out the other side. That is exactly what this library is for. Grab a transparent PNG and convert it to WebP to confirm the transparency survives. Take a CMYK TIFF into a JPG and check the colors. Run a DOCX to PDF, an XLSX to CSV, or a HEIC to JPG, and see whether the result is what you expected.

The pairing works because the samples are chosen to surface the differences that matter between formats rather than to look nice. Transparency, color profiles, resolution, and container behavior are precisely the things conversions get wrong, and they are precisely the things the image and document samples are built to probe. If you want a guided run through the converter itself, How to convert and compress files privately with Novus Convert walks through a first conversion, and you can bring any of these files along as the input.

It also cuts out a step you would otherwise repeat constantly. Instead of exporting a one-off file from some other application every time you want to test a conversion, you download a known, consistent sample, run it through Novus Convert, and compare against a stable baseline. When the input never changes, a difference in the output is a signal about the converter rather than noise from a sample you cobbled together differently each time.

The other tool Novus Examples was built to feed is the NSS Background Remover at bgremover.novusstreamsolutions.com. Background removal is unusually sensitive to the kind of image you give it, and the images that expose its limits are the ones most people never think to test with. A clean product shot on a plain backdrop is easy. The hard cases are the ones with fine detail at the edges, and those are the cases the library keeps ready.

That is why the image catalog includes hair and edge-heavy portraits, transparent PNGs, and grayscale samples specifically for this kind of testing. Wispy hair against a busy background is the classic thing that separates a careful cutout from a crude one, so having a portrait built to stress that boundary lets you judge an edge honestly rather than on a photo that happened to be easy. Transparent inputs let you check how the tool behaves when part of the image already has no background, and grayscale samples confirm it does not quietly assume every image arrives in full color.

The point, as with the converter, is a stable and deliberately difficult input. When you always test against the same demanding portrait, you can tell whether a result is genuinely good or merely lucky, and you can compare one attempt to the next on equal terms. If you are new to the tool, How to remove a background from an image with NSS Background Remover covers a first removal, and any of the portrait samples make a fair, repeatable test image to bring to it.

Most sample libraries stop at well-formed files, and that is where the interesting testing actually begins. Software rarely fails on the clean, ordinary input it was demonstrated with. It fails on the empty upload, the file that claims to be a PNG but is not, the archive that was truncated in transit, and the video that is simply too large for the limit you set. Those are the moments that decide whether your tool degrades with a clear message or falls over with a stack trace, and they are the moments you cannot rehearse without the awkward files to rehearse against.

So Novus Examples keeps a specialty section for exactly these cases. There are zero-byte files for confirming your code handles an empty input rather than assuming every upload has content. There are deliberately corrupt and malformed files for checking that a parser rejects bad data cleanly instead of crashing or, worse, half-accepting it. There are wrong-extension files, where the name and the actual contents disagree, for testing whether your validation trusts the extension or looks at the bytes. And there are oversized files for exercising the size limits and timeouts that only ever get tested in production if you skip them here.

Error handling is the part of a product users only notice when it is missing, and it is genuinely hard to test on purpose because the inputs that trigger it are inconvenient to produce. Making those inputs a single download is the whole reason this section exists. Robustness is not a feature you can bolt on at the end; it is something you confirm case by case, and having the difficult cases ready to hand is what turns robustness testing from an intention into a habit.

Novus Examples is free, it asks for no account, and the files download straight in your browser. There is no sign-up wall, no email to hand over, and no metering that counts how many samples you have taken. You open the site, find the format you need, and download it, in the same few seconds it takes to describe. A library of test files should be the least ceremonious thing in your day, and we built it to behave that way.

This is the same stance that runs through the rest of the portfolio, applied to sample files rather than to processing media. There is no reason a page whose entire job is to hand you a file should collect anything about you in return, so it does not. The tool respects your privacy by doing less: it serves the samples and gets out of the way, and there is nothing about your visit that needs to be stored to make that work.

Being free is a natural state here rather than a promotion we might later walk back. Serving a catalog of static sample files does not carry a per-use bill quietly accruing in the background, which means keeping Examples free is a straightforward commitment rather than a subsidy waiting to be reclaimed behind a paywall. We would rather build the tool so that free is simply what it is than launch generously and reconsider once anyone starts to depend on it.

Novus Examples is the sixth live app in the portfolio, joining the NSS Background Remover, Novus Visualizers, Novus PDF Studio, Novus Convert, and Novus Learn. The through-line across all six is the same one it always has been: build free tools that hold their own against paid alternatives, keep the work close to the person using it, and never ask for more than the tool needs. Examples fits that line from a slightly different angle. It is not the tool you run your file through; it is the file you run through the tool, and it makes the other apps easier to trust because you can test them with inputs you actually chose.

You can open it at examples.novusstreamsolutions.com and browse straight to the category you need, from images and data to documents, audio, video, source code, and security, or search and browse by the job you are testing for. For the full picture of how the catalog is organized and what each sample is meant to exercise, see Novus Examples, and for the current inventory of every category, see Tool maps. If you would like a guided first pass, Getting started with Novus Examples: find and download a sample file walks through picking a format, downloading a sample, and running it through the tool you are trying to test. It is a small tool for a small, recurring problem, and that is exactly why we think it earns its place: the next time your work is blocked on the want of a suitable file, the file is already here.

As of the July 31, 2026 release gate the catalog stands at 2,558 files across 17 categories and 221 distinct extensions, and the per-category counts are worth stating because they show where the work actually went: Images 452, Data 423, Video 401, Audio 154, Documents 152, Templates 146, Forms 134, Source code 125, AI/ML 124, Localization 90, Web assets 81, Models 64, Security 58, Archives 56, Email 50, E-books 31, Fonts 17.

Images and Data being large is not an accident. They are the categories where a tool is most likely to be subtly wrong rather than obviously broken (a colour space silently shifted, an alpha channel ignored, a CSV quoting rule mishandled) and those failures need a wide spread of fixtures to catch.

Video is the newest reason the total moved. It was 70 files and the thinnest part of the catalog (the most subcategories of any category and the fewest files in each) and it is now 401 and the third largest. The full account is in Novus Examples grows up: 401 video fixtures, 480 templates, and OBS-compatible scenes, along with the visual template library and the OBS-compatible scene collections that came with it. That library has since grown again, to 504 editable visual templates across 42 categories, alongside 49 catalog-backed form definitions and 43 curated document families, 36 of which open directly in Document Studio.

A format taxonomy is how a library is organised and not how testing works. Nobody wakes up needing a TIFF; they wake up needing something that will break their OCR pipeline. The /for routes invert the catalog so you start from the job (OCR, ASR, colourization, form testing, CSV parsing, visual diff and regression, segmentation, schema validation, image enhancement, PDF editor testing, JWT and JWKS, background removal) and get fixtures pulled from wherever they happen to live.

The other addition covers what a downloadable file fundamentally cannot test. Twelve live targets at /targets are real pages, each exhibiting exactly one behaviour: ARIA landmarks, nested modals with focus trapping and Escape stacking, infinite scroll with a sentinel, a print stylesheet, a slow-loading page, a cookie banner, a sortable and filterable table, a file drop zone that uploads nothing, a nested iframe chain, a scroll-spy nav, and pages that force dark and force light.

The isolation is the feature. A real site does a dozen things at once and changes without warning, so a failed test against one tells you very little. When a scraper stalls on a page whose only job is to append content behind a sentinel, the cause is not in question.