Every file is tagged with what it's good for testing. Pick a use case to see its fixtures — or filter the full library.
Images & vision
Product-on-background compositions, alpha PNGs over transparency, and hard-case edges — fur, mesh, glass refraction, colour-matched backgrounds — for scoring removers and recomposite workflows.
Images & vision
Bar, line, pie, and scatter charts rendered as images with labelled axes, legends, and known values — for testing chart-extraction and chart-to-data tools, vision models, and OCR of embedded text.
Images & vision
Tagged versus untagged images, ICC-profiled samples, and CMYK/wide-gamut cases — for testing colour-management pipelines, profile handling, and conversions that must not assume sRGB.
Images & vision
True-colour ground truths paired with greyscale and sepia inputs for measuring photo colourization models and filters against known colour references.
Images & vision
The same source saved at quality 10/50/90 for comparing compression artefacts and quality settings.
Images & vision
Scenes shipped with the structure-control inputs a guided-generation model consumes — Canny, depth, normal, segmentation, scribble, line-art, and pose — each derived from the same colour ground truth so outputs can be scored against their source.
Images & vision
Clean references with gaussian, motion, and camera-shake blurred variants — documented blur type and strength — for measuring deblur and sharpening accuracy with PSNR/SSIM against ground truth.
Images & vision
Noisy images paired with their exact clean reference, with documented noise type, sigma, and seed — everything you need to measure how well a denoise filter recovers the original.
Images & vision
Synthetic depth maps with a known near/far layout and matching colour reference, for depth-conditioned generation and monocular depth-estimation regression.
Images & vision
A/B image twins with controlled pixel changes, anti-alias variants, and UI chrome crops for screenshot diff and visual regression tools.
Images & vision
Binary Canny-style edge maps paired with the exact image they were extracted from — measure an edge detector or edge-conditioned generator against a documented reference.
Images & vision
Clean references with underexposed, overexposed, low-contrast, cast, haze, and heavy JPEG-artefact variants for auto-enhance, dehaze, and artefact-repair tools.
Images & vision
Images with embedded EXIF orientation flags for testing whether your tool respects rotation metadata.
Images & vision
Synthetic face illustrations (not real people) with JPEG crush, blur, and downscale damage plus masks — for testing face restore and enhancement tools against clean ground truth.
Images & vision
Colour originals paired with true-greyscale conversions, grey wedges, and gradient ramps for testing desaturation and tone handling.
Images & vision
16-bit linear TIFF ramps and twins (not PQ/HDR10 display masters) for deep-colour loaders, truncation, and tone-map pipelines.
Images & vision
Orientation ladders, SAMPLE GPS EXIF, and sRGB vs Adobe RGB tagged images for colour-managed pipelines and EXIF readers.
Images & vision
Subject images paired with soft-alpha mattes and trimaps (fur, mesh, glass edges) — for testing image-matting models and alpha extraction against a known ground truth.
Images & vision
Deterministic images with known properties for regression-testing resize, crop, convert, and filter pipelines.
Images & vision
RGB scenes paired with class or instance masks and documented class IDs and bounding boxes — for scoring semantic and instance segmentation against a known ground truth.
Images & vision
Images with cut-out regions (rects, circles, corners, strips, irregular tears) and matching masks for testing inpainting, generative fill, content-aware fill, and object-removal pipelines.
Images & vision
Clean line-art extractions (black lines on white) paired with their colour source — for line-art-conditioned generation and line-extraction quality tests.
Images & vision
Hair/fur-like soft alpha composites and trimap-style inputs for matting and refined background-removal tools.
Images & vision
A dense ground-truth mesh plus a low-poly simplification of the same shape, with triangle counts recorded — score a decimation / LOD tool by comparing the two.
Images & vision
Deterministic GLB meshes with known geometry and triangle counts for regression-testing mesh import/export, decimation, repair, and validation pipelines.
Images & vision
A watertight mesh paired with a deliberately broken twin (holes, missing faces, open ends) — the input and intended result for a mesh-repair / hole-filling tool.
Images & vision
Tangent-space surface-normal maps encoded from a known height/depth field — inputs for normal-conditioned generation, relighting, and normal-estimation checks.
Images & vision
Image-only 'scanned' documents paired with their text source, for measuring OCR accuracy against a known ground truth.
Images & vision
Full-frame references with empty-border inputs and expand masks for testing outpainting, generative expand, and content-aware extend tools.
Images & vision
Complete PBR material sets — base colour, normal, roughness, metallic, ambient occlusion, and height — for testing material-generation models and physically-based renderers channel by channel.
Images & vision
Surface scans shipped as clean ground truth plus a degraded capture, for scoring photogrammetry and surface-reconstruction pipelines against a known target.
Images & vision
A dense, noise-free point cloud paired with a sparse, noisy scan of the same surface — a reconstruction / denoise target for point-cloud tooling.
Images & vision
OpenPose-style keypoint skeletons for figure subjects, paired with the rendered figure — a pose-control input for guided generation and a target for pose estimators.
Images & vision
Images from tiny to 4K, plus vertical and extreme aspect ratios, for testing scaling and layout.
Images & vision
Clean references paired with damaged inputs and damage masks — creases, scratches, dust, water stains, torn corners, and cuts — for measuring photo restoration and damage-repair tools against ground truth.
Images & vision
Loose, hand-drawn-style scribble maps derived from a scene's edges — the sparse control input for scribble-guided image generation.
Images & vision
Multi-class RGB scenes with indexed masks and class maps for semantic/instance segmentation loaders and eval harnesses.
Images & vision
Semantic and instance segmentation maps paired with their source scene — the label reference for segmentation-conditioned generation and mask-quality scoring.
Images & vision
Content photos paired with posterize, line-art, paint-like, comic, and geometric style references for neural and filter style-transfer pipelines.
Images & vision
Images with isolated foreground subjects, cluttered backgrounds, and exact masks for testing saliency, subject selection, cropping, and background-removal pipelines.
Images & vision
Large clean references with deliberately downscaled (and optionally JPEG-degraded) inputs for testing upscalers, super-resolution models, and enlarge-without-blur pipelines.
Images & vision
Tileable texture channels with documented roles and resolution, for validating texture pipelines, atlas packers, and material import/export.
Images & vision
A resolution ladder and varied aspect ratios for exercising thumbnail and preview generators.
Images & vision
Clean images paired with synthetic SAMPLE watermarks (logo, text, tile, translucent) and masks — for scoring watermark removers without using real brand marks.
Audio & video
VAST ad responses and SCTE-35 splice descriptors covering linear break signalling, wrappers, and companion creatives — for testing ad-decision integration and manifest-manipulation logic without calling a real ad server.
Audio & video
Short synthetic digit/tone utterances with transcript JSON and clean↔noise pairs — for testing ASR loaders, WER harnesses, and audio preprocessing.
Audio & video
Pure tones, sweeps, and noise with known frequencies and levels for testing spectrum, level, and waveform tools.
Audio & video
Synthetic foley, ambience, dialogue, loops, stems, cue sheets, and edge cases with documented timing and signal properties for repeatable editing checks.
Audio & video
Tones with precisely documented leading and trailing silence timestamps — direct fixtures for auto-trim tools.
Audio & video
The same cue list written with a UTF-8 BOM, without one, in UTF-16, with CRLF and with bare LF, plus non-Latin scripts and right-to-left text — for finding the parser that assumed captions are always LF-terminated ASCII.
Audio & video
Documented loudness ladders, clipped vs clean twins, and sample-rate conversion sets for loudness meters and normalizers.
Audio & video
Containers carrying several subtitle and audio tracks at once, with real language tags and the default, forced, hearing-impaired and visual-impaired flags actually set — for testing track pickers, accessibility menus, and the flag handling that only shows up when more than one track exists.
Audio & video
Video files with embedded chapters, full container tags, attached cover art and a SMPTE start timecode — each paired with a deliberately stripped twin, so you can tell a metadata reader that found nothing from one that was never given anything.
Audio & video
RSS, Atom, and JSON Feed syndication, an OPML subscription list, and M3U / HLS M3U8 media playlists — for testing feed readers, podcast apps, and media players.
Audio & video
Small synthetic sound-effect hits, room tones, layers, and loopable material with exact duration and level contracts for sound-design and media-pipeline QA.
Audio & video
HLS playlists and DASH MPDs covering multi-bitrate ladders, byte-range addressing, and segment templates. Served inline with permissive CORS under /files/stream/, so a player can load them from any origin without re-hosting.
Audio & video
The same captions written out across SubRip, WebVTT, ASS/SSA, SBV, TTML and a synced LRC lyric file, with known timings — for exercising subtitle parsers, format converters and burn-in tools against every serialisation.
Audio & video
SubRip and WebVTT sidecars plus multi-track containers — forced-narrative, SDH and styled tracks in one file — for testing track selection, default flags and caption rendering in players.
Audio & video
Equivalent deterministic edit decisions and project structures across common timeline formats, with expected clip order, timing, transitions, and failure behavior.
Audio & video
One source encoded as H.264, HEVC, AV1, VP9, MJPEG, and a ProRes-compatible mezzanine, across 8/10/12-bit and 4:2:0/4:2:2/4:4:4 — for testing decoder support, transcode pipelines, and container remuxing.
Audio & video
Greyscale, sepia and partially-desaturated clips shipped beside the full-colour original they were derived from, including a 24-patch chart with documented values — so a colourisation model's output can be scored per patch rather than judged by eye.
Audio & video
Motion-blurred and defocused clips produced from a sharp source with a recorded kernel, so deblurring output can be measured against the original rather than compared to another blurred frame.
Audio & video
Clips degraded with documented noise type, sigma, and seed, each shipped beside the exact high-quality source it came from — so a denoiser's output can be scored, not eyeballed.
Audio & video
Layered scenes whose planes translate at documented, fixed ratios, so relative depth is defined by measurable motion rather than inferred from pictorial cues — giving depth-from-video output an objective reference.
Audio & video
Short playable clips plus overlays, captions, timelines, colour targets, sync references, and controlled failures for testing browser and desktop editing workflows.
Audio & video
Clips with frames removed on a known pattern, paired with the full-rate source — for scoring frame-interpolation and slow-motion tooling against a real reference.
Audio & video
Moving subjects over known backgrounds, shipped with the exact alpha matte used to composite them — for scoring background removal and green-screen keying frame by frame.
Audio & video
Short 270p clips with known scene cuts, freezes, black frames, flash frames, and burned-in captions — for detection and QC tooling.
Audio & video
The same footage at a ladder of quantiser settings from near-lossless down to visibly broken, each paired with its high-quality source — for testing artefact-removal models and for calibrating quality metrics against known encoder settings.
Audio & video
Hard-edged and anti-aliased subjects in motion, each shipped with the exact per-frame mask used to create it — for scoring video object segmentation and rotoscoping frame by frame instead of spot-checking.
Audio & video
Camera shake applied with a recorded motion path over a static original, so a stabiliser's residual motion can be measured rather than estimated.
Audio & video
Low-resolution clips produced from a documented high-quality source by a recorded filter, so super-resolution output can be measured against the original rather than judged by eye.
Forms & documents
Fillable PDF forms with text, choice, checkbox, and radio fields for AcroForm tooling.
Forms & documents
Realistic HTML forms with labelled fields (tax, KYC, shipping, insurance, legal, finance) for exercising browser autofill and form-field detectors.
Forms & documents
Text-based files you can open, edit, and download directly in the browser editor.
Forms & documents
PDF AcroForms with documented fields and live HTML forms for testing form parsers and fillers.
Forms & documents
HTML and PDF form files for testing scrapers, autofill, validators, and PDF form fillers across real-life use cases.
Forms & documents
Form PDFs, bookmarked documents, scanned pairs, and deliberately corrupt files for exercising PDF editors, parsers, and fillers.
Forms & documents
AcroForm PDFs and FDF/XFDF data twins for testing PDF form fillers and form-data import.
Forms & documents
Multi-sheet workbooks with documented formulas and plain sheets for testing spreadsheet parsers and importers.
Data & code
OpenAPI/Swagger specs, GraphQL SDL, JSON Schema, paginated and problem+json error payloads, and webhook samples — for testing API clients, mock servers, contract tests, and schema validators.
Data & code
iCalendar (.ics) files with events, timezones, and recurrence — for testing calendar imports and parsers.
Data & code
Short, known-correct source files in many languages with classes, functions, generics, enums, and error handling — for testing parsers, linters, formatters, language detection, and diff viewers.
Data & code
TOML and INI configuration files with nested sections and typed values — for testing config parsers and loaders.
Data & code
TOML, INI, YAML, .env, and dotfile configuration samples with nested sections and typed values — for testing config parsers, loaders, and environment tooling.
Data & code
vCard (.vcf) contact files with names, emails, phones, and addresses — for testing contact importers.
Data & code
Clean and deliberately messy CSVs — quoted commas, embedded newlines, ragged rows, odd delimiters, and encodings.
Data & code
Columnar (Parquet/ORC/Feather), row (Avro), and messaging (MessagePack/CBOR/Protobuf) formats plus star-schema and log data — for testing ETL, data-lake ingestion, and warehouse loaders.
Data & code
Realistic faker-generated datasets with documented schemas for testing import and ETL flows.
Data & code
Standards-compliant RFC 822 messages — plain, multipart text+HTML, and with an attachment — plus an MBOX mailbox, for testing header parsing, MIME decoding, attachment extraction, and mailbox splitting.
Data & code
UTF-8, UTF-8-BOM, UTF-16, and Latin-1 files with documented encodings and line endings for testing charset detection.
Data & code
Well-formed RSS 2.0 and Atom feeds with multiple entries — for testing feed readers and parsers.
Data & code
GeoJSON, GPX, and KML files with points, lines, polygons, and tracks — for testing map tools, route parsers, and geo importers.
Data & code
Node/edge datasets in GraphML and GEXF (directed and undirected, with attributes and weights) — for testing network importers, layout tools, and graph converters.
Data & code
Standalone documents, email markup, forms, malformed edge cases, and semantic page structures for testing HTML parsers, extractors, sanitizers, and DOM-import pipelines.
Data & code
Flat, deeply nested, JSON Lines, and intentionally invalid JSON for testing parsers and error handling.
Data & code
Access logs and JSON-lines application logs — for testing log parsers, tailers, and ingestion pipelines.
Data & code
Structured and plain-text telemetry with known timestamps, levels, request identifiers, and error states for testing log ingestion, correlation, dashboards, and alert pipelines.
Data & code
Documented size, row-count, duration, and resolution ladders for measuring parser, renderer, converter, and upload performance without relying on private production data.
Data & code
Valid and intentionally invalid OpenAPI/JSON Schema documents plus request/response examples for schema validators and API tooling.
Data & code
JSON Schema documents describing a data shape — for testing validators and schema-aware tooling.
Data & code
Citation catalogs (BibTeX, RIS), chemistry structures (MDL Molfile, PDB), and gridded binary data (NetCDF, FITS) — for testing reference managers, molecule viewers, and scientific-data loaders.
Data & code
Known records represented as JSON, MessagePack, CBOR, BSON, Protobuf, Arrow, and related formats for testing round trips, schema handling, type preservation, and binary decoders.
Data & code
Postgres- and MySQL-flavour SQL dumps with CREATE TABLE and INSERT statements — for testing SQL importers and migrations.
Data & code
Idiomatic hello-world programs and realistic snippets across sixteen languages, exercising comments, string escapes, interpolation, numeric literals, and language keywords — for testing syntax highlighters, editor themes, and tree-sitter grammars.
Data & code
Sensor, market, event, and telemetry series with documented intervals, gaps, duplicates, and timezone behavior for testing importers, resampling, charts, and anomaly pipelines.
Data & code
Irregular timestamps, DST gaps, and duplicate keys for time-series importers and charting libraries.
Data & code
Favicons, web app manifests, service workers, robots and sitemap files, Open Graph images, and .well-known resources — for testing web tooling, crawlers, PWA installers, and asset pipelines.
Data & code
Semantic pages, tables, forms, feeds, metadata, and downloadable fixtures for testing extraction and browser automation against stable, purpose-built targets.
Conversion & robustness
The same content exported across many formats and linked as a group, so you can convert one and diff against the expected twin.
Conversion & robustness
Deliberately corrupt and invalid files, clearly labelled, for testing how your tool fails.
Conversion & robustness
Prompt plans, input media, expected reference outputs, manifests, hashes, and controlled failures for exercising ingest, transform, validate, and export stages together.
AI / ML
Ordered creator briefs, prompt plans, tool-shaped messages, media inputs, reference outputs, and controlled edge cases joined into complete six-artifact workflow families.
AI / ML
A rendered detection scene annotated in COCO, YOLO, and Pascal-VOC formats — for testing annotation loaders, format converters, and vision pipelines against a known image.
AI / ML
The same texts as a normalised vector set in JSON, Parquet, and NumPy .npy — for testing vector stores, similarity search, and embedding loaders.
AI / ML
Labelled, synthetic datasets in the shapes ML pipelines expect — JSONL for text tasks, image annotations, embeddings, and sample weights — for testing data loaders, tokenizers, and training tooling.
AI / ML
Benchmark results, confusion matrices, ROC curves, and classification reports in CSV and JSON — for testing eval dashboards, metric parsers, and leaderboard importers.
AI / ML
Small, validated ONNX graphs paired with named JSON inputs and expected tensor outputs. Use them to test runtime loading, dtype and shape handling, broadcasting, dynamic batches, reduction behavior, and numeric tolerances without downloading a production-sized model.
AI / ML
Sentiment, NER, chat, instruction-tuning, QA, summarization, and translation data in JSON Lines and JSON — for testing NLP loaders, tokenizers, and format converters.
AI / ML
Synthetic scenes with exact class labels, bounding boxes, and matching COCO, YOLO, or Pascal VOC annotations for testing detection loaders, converters, and evaluation code.
AI / ML
A templated prompt library in JSON Lines with tasks, tags, and placeholders — for testing prompt-management tools and JSONL parsers.
AI / ML
Vendor-neutral plans and provider-compatible message fixtures paired with explicit inputs, assertions, failure cases, and expected results for repeatable prompt regression testing.
AI / ML
JSONL traces of function/tool calls with arguments and results — for testing agent harnesses and tool routers.
Security & privacy
Self-signed X.509 certificates (PEM, CRT, DER), a CSR, RSA and Ed25519 keys, an SSH public key, a PKCS#12 bundle, and an htpasswd file — all published sample-only material, for testing certificate parsers, TLS tooling, keystore importers, and PEM/DER decoders.
Security & privacy
Unsigned and SAMPLE-signed JWT variants plus JWKS documents — published sample material only, for auth parser tests.
Security & privacy
Images, audio, video, and documents with documented metadata paired with deliberately stripped versions, for verifying extraction, preservation, redaction, and privacy-scrubbing behavior.
Security & privacy
Sample subject-access ZIP packages, DPA/consent templates, and fictional PII — for testing privacy-export parsers and consent UX tooling.
Localization & templates
Compact prompt, audio, and video artifacts for testing a creator pipeline from brief and planning through editing, evaluation, and final export.
Localization & templates
Gettext PO/POT, XLIFF, Apple .strings, Flutter .arb, Android strings.xml, .NET .resx, and i18next JSON — with RTL and CJK variants — for testing localization pipelines, translation-memory tools, and catalog converters.
Localization & templates
Parallel text and accented, multi-script content — for testing translation pipelines, Unicode handling, and localization tooling.
Localization & templates
Professional, neutral starting-point documents you can download and adapt — invoices, resumes, budgets, and more.
We use Google Analytics and show ads via Adsterra. Non-essential cookies and ad scripts run only after you allow the matching categories. See our cookie policy.
