| --- |
| title: README |
| emoji: 🗂️ |
| colorFrom: blue |
| colorTo: green |
| sdk: static |
| pinned: false |
| --- |
| |
| # Kled |
|
|
| **Consented, commercially licensed training data from 625,000+ contributors in 100+ countries, strongest in Southeast Asia, the United States and Europe.** |
|
|
| Kled is a consumer app. Contributors complete paid data tasks on their own phones, in their own homes, streets, kitchens, workplaces and languages, and opt in to commercial use when they upload. Each submission is reviewed against the task brief before it enters a dataset. The result is data that public benchmarks and scraped corpora do not have: camera-original phone photos with EXIF intact, native PDFs of real finances, third-person video of real chores, speech in real noise, scams as they actually arrived, screens as people actually see them. |
|
|
| The 127 dataset cards in this organization describe what each set contains, how it was captured, what is labeled, how it compares with the public alternatives and how to license it. Files are delivered to licensees by Kled, not hosted here. Every card has a **Request a sample set** form at the top; the data team also answers at [team@kled.ai](mailto:team@kled.ai). |
|
|
| ## How licensing works |
|
|
| 1. **Sample set.** Request one from any dataset card or from the catalog at [kled.ai/datasets](https://www.kled.ai/datasets?utm_source=huggingface&utm_medium=org_card). |
| 2. **Scope.** Tell the data team the volume, geographies, languages, annotation and exclusivity you need. Custom collection on the same contributor base is available for most tasks. |
| 3. **License.** Non-exclusive commercial training license by default; exclusivity on request. Every contributor opted in to this use when they uploaded. |
|
|
| ## Collections |
|
|
| - [Start here: Kled flagship datasets](https://huggingface.co/collections/useKled/start-here-kled-flagship-datasets-6ade92455cf87498192affd0) (18 datasets): The largest and most requested sets in the catalog. Every file is phone-captured by a consenting contributor and licensed for commercial training. |
| - [Robotics and embodied AI video](https://huggingface.co/collections/useKled/robotics-and-embodied-ai-video-6ade92a1277e5ce217b0145e) (31 datasets): Hands, tools, objects and rooms on phone video: pick-and-place, articulated objects, deformables, cooking, cleaning, repair, navigation. |
| - [Fraud, scam and synthetic-media detection](https://huggingface.co/collections/useKled/fraud-scam-and-synthetic-media-detection-6ade92ba0b55f23a90bacf4b) (15 datasets): Real scam messages, phishing screenshots, fake apps, forged documents, altered receipts, counterfeits and AI images as they reach people. |
| - [GUI, web and app agents](https://huggingface.co/collections/useKled/gui-web-and-app-agents-6ade92cb26d7d158c7aa4229) (10 datasets): Screenshots and screen recordings of real apps, settings, errors, notifications, checkout and booking flows and government portals. |
| - [OCR, documents and scene text](https://huggingface.co/collections/useKled/ocr-documents-and-scene-text-6ade924ee273bf4e02893bef) (29 datasets): Receipts, bills, labels, menus, signs, packaging, forms, native PDFs, handwriting and non-Latin scripts, contributed by their owners. |
| - [Southeast Asia](https://huggingface.co/collections/useKled/southeast-asia-6ade92d71d2d900a5a6eb7cd) (28 datasets): Sets collected across the 11 countries of Southeast Asia, where Kled's contributor base is strongest: Indonesia, Malaysia and the Philippines. |
| - [Kled: Real-World Capture](https://huggingface.co/collections/useKled/kled-real-world-capture-6ade926a7384165132d1a30d) (54 datasets): Phone photos, screenshots and short video of streets, shelves, screens, scams, weather and packaging as contributors meet them. |
| - [Kled: Tools in Use](https://huggingface.co/collections/useKled/kled-tools-in-use-6ade9283ce42e99bf48a7f6d) (24 datasets): Video of people doing ordinary tasks with their hands: cooking, folding, repairing, sorting, cleaning. For robot learning and world models. |
| - [Kled: Financial & Legal Documents](https://huggingface.co/collections/useKled/kled-financial-and-legal-documents-6ade928e2bd149d417f968e1) (14 datasets): Native PDFs and photographed paper from real finances and contracts: statements, invoices, payslips, leases, receipts, tax forms. |
| - [Kled: Everyday Writing](https://huggingface.co/collections/useKled/kled-everyday-writing-6ade92956b9f6c51122a9050) (13 datasets): Text people wrote for their own reasons, contributed by the author: emails, notes, reviews, recipes, how-to guides, support threads, code. |
| - [Kled: Research & Education](https://huggingface.co/collections/useKled/kled-research-and-education-6ade923d6b9f6c51122a846e) (11 datasets): Handwriting at scale, sketches, whiteboards, field notes, technical drawings and online pen strokes, for OCR, math and reasoning work. |
| - [Kled: Speech & Audio](https://huggingface.co/collections/useKled/kled-speech-and-audio-6ade929c1d2d900a5a6eb152) (9 datasets): Phone-recorded speech: read phrases, spoken digits and addresses, accents, code-switching, low-resource SEA languages, conversations. |
| - [Kled: Healthcare & Clinical](https://huggingface.co/collections/useKled/kled-healthcare-and-clinical-6ade9243a902f78ade615268) (2 datasets): De-identified intake forms and imaging studies, scoped with the buyer and reviewed before release. |
|
|
| ## The catalog |
|
|
| ### Real-World Capture (54) |
|
|
| Phone photos, screenshots and short video of the world as contributors meet it: streets, shelves, screens, scams, weather, packaging and the long tail no survey fleet covers. |
|
|
| | Dataset | What it is | Volume | |
| |---|---|---| |
| | [AI Images in the Wild (250,000 screens)](https://huggingface.co/datasets/useKled/ai-generated-images-found-in-the-wild) | Screenshots of images contributors judged AI-generated on social platforms, with platform, source, reason, confidence. | 250,000+ screenshots | |
| | [Six-View Product Photos (10,000 products)](https://huggingface.co/datasets/useKled/all-six-faces-of-one-package) | Six-photo sets of one retail product: front, back, both sides, top and a barcode close-up, same lighting throughout. | 10,000+ products | |
| | [App Error Screens (10,000 screens)](https://huggingface.co/datasets/useKled/app-error-screens) | Full-screen screenshots of live app and site failures (spinners, "something went wrong", payment failed), in context. | 10,000+ screenshots | |
| | [Inner App Settings (100,000 screens)](https://huggingface.co/datasets/useKled/app-settings-screens) | Settings pages one level deep inside real apps, every toggle, slider and option visible. Top-level menus excluded. | 100,000+ screenshots | |
| | [Naturally Failed Photos (150,000 photos)](https://huggingface.co/datasets/useKled/bad-photos-you-d-normally-delete) | Photo failures people would have deleted: motion blur, missed focus, tilted horizons, cut-off subjects. Real phones. | 150,000+ photos | |
| | [SEA Construction Sites (5,000 photos)](https://huggingface.co/datasets/useKled/building-sites) | Street-level photos of Southeast Asian building sites focused on a prompted element, no faces, hazards noted. | 5,000+ photos | |
| | [Website Flow Screenshots (5,000 flows)](https://huggingface.co/datasets/useKled/checkout-step-by-step) | Ordered 4 to 6 screenshot sequences of one real website flow (sign-up, basket, booking, settings), each step described. | 5,000+ flows | |
| | [Verified Object Counts (25,000 images)](https://huggingface.co/datasets/useKled/cluttered-surfaces-for-counting) | Overhead photos of cluttered surfaces with ten or more objects, each with an exact count entered by the photographer. | 25,000+ images | |
| | [Counterfeit and Genuine Pairs (5,000 pairs)](https://huggingface.co/datasets/useKled/counterfeit-vs-authentic-products) | Eight-photo sets pairing a fake and a genuine version of one product, same light, same four angles, which is fake noted. | 5,000+ pairs | |
| | [Home Damage, Three Frames (10,000 sets)](https://huggingface.co/datasets/useKled/damage-in-your-home) | Three-shot sets of household damage (damp, cracks, leaks, peeling paint) from whole room to close-up, defect type named. | 10,000+ photo sets | |
| | [Damaged Parcels, Three Frames (5,000 sets)](https://huggingface.co/datasets/useKled/damaged-deliveries) | Three-photo sets of damaged deliveries in Southeast Asia: whole item, damage close-up, packaging with its label. | 5,000+ photo sets | |
| | [Fake Delivery Notice Scams (5,000 screens)](https://huggingface.co/datasets/useKled/fake-delivery-notices) | Fake "missed delivery" texts and emails in Southeast Asia, posing as regional couriers, sender and link in frame. | 5,000+ screenshots | |
| | [Fake Trading App Screens (5,000 screens)](https://huggingface.co/datasets/useKled/fake-trading-app-screens) | Fraudulent trading platforms, fake exchange dashboards and recruitment messages in Southeast Asia, scam type noted. | 5,000+ screenshots | |
| | [Near-Identical Object Groups (5,000 images)](https://huggingface.co/datasets/useKled/fine-grained-similar-objects) | Close-ups of three or more almost identical items in one frame, each with the owner's note on what tells them apart. | 5,000+ images | |
| | [Nutrition Label Panels (50,000 photos)](https://huggingface.co/datasets/useKled/food-labels) | Square-on photos of nutrition facts and ingredient panels from food sold in 11 Southeast Asian countries, rows legible. | 50,000+ photos | |
| | [Camera-Original Phone Photos (1M photos)](https://huggingface.co/datasets/useKled/fresh-photos-taken-right-now) | Unedited photos shot inside the Kled camera, EXIF intact, no filters or camera-roll imports. Global contributor base. | 1,000,000+ photos | |
| | [SEA Government Portals (300,000 screens)](https://huggingface.co/datasets/useKled/government-service-websites) | Full-page screenshots of public information and form pages on government sites across 11 Southeast Asian countries. | 300,000+ screenshots | |
| | [Hand-Painted SEA Signage (5,000 photos)](https://huggingface.co/datasets/useKled/handwritten-public-signage) | Straight-on photos of handwritten and hand-painted signs in Southeast Asian shops, markets and streets. | 5,000+ photos | |
| | [Impersonation Scam Chats (10,000 screens)](https://huggingface.co/datasets/useKled/impersonation-scam-messages) | Whole-chat screenshots of scammers posing as a bank, employer or company, claimed identity and the ask both visible. | 10,000+ screenshots | |
| | [Scam Marketplace Listings (5,000 listings)](https://huggingface.co/datasets/useKled/marketplace-scam-listings) | Full screenshots of suspected scam listings on Shopee, Lazada, Facebook Marketplace and peers, red flags noted. | 5,000+ listings | |
| | [Mobile App Screenshots (200,000 screens)](https://huggingface.co/datasets/useKled/mobile-app-screens) | Whole-screen screenshots of live apps, status bar and navigation included, on contributors' own phones, by screen type. | 200,000+ screenshots | |
| | [Non-English Packaging Text (50,000 photos)](https://huggingface.co/datasets/useKled/non-english-product-packaging) | Close, square-on photos of consumer packaging whose main text is not English, from 11 Southeast Asian countries. | 50,000+ photos | |
| | [Non-Latin Scene Text (25,000 photos)](https://huggingface.co/datasets/useKled/non-latin-script-text) | Sharp photos of real-world writing in Devanagari, Arabic, Thai, Cyrillic, Han and other non-Latin scripts, tagged. | 25,000+ photos | |
| | [Notification Shade Stacks (150,000 screens)](https://huggingface.co/datasets/useKled/notification-stacks) | Screenshots of notification shades and lock screens with two or more notifications stacked, status bar included. | 150,000+ screenshots | |
| | [Parcel Shipping Labels (25,000 labels)](https://huggingface.co/datasets/useKled/parcel-shipping-labels) | Straight-on photos of shipping labels on real parcels, barcodes and small print sharp, from couriers in many countries. | 25,000+ labels | |
| | [Smishing Screenshots (50,000 screens)](https://huggingface.co/datasets/useKled/phishing-text-messages) | Scam text messages as received in 11 Southeast Asian countries, sender line and full message intact. | 50,000+ screenshots | |
| | [Low-Light Phone Photos (150,000 photos)](https://huggingface.co/datasets/useKled/photos-in-the-dark) | Photos taken at night or in dim rooms with the flash off and no brightening afterward, straight from the Kled camera. | 150,000+ photos | |
| | [SEA Plant Health Close-Ups (50,000 photos)](https://huggingface.co/datasets/useKled/plants-close-up) | Close-ups of healthy and unhealthy plants in 11 Southeast Asian countries, leaf and stem detail, grower's note attached. | 50,000+ photos | |
| | [SEA Menus and Price Boards (10,000 photos)](https://huggingface.co/datasets/useKled/restaurant-menus) | Whole-page photos of menus, carinderia boards and warung price lists in 11 Southeast Asian countries, prices readable. | 10,000+ photos | |
| | [Same Screen, Three Modes (5,000 sets)](https://huggingface.co/datasets/useKled/same-screen-different-display-settings) | Controlled triplets of one app screen as it normally looks, in dark mode, and at maximum system text size. | 5,000+ screen sets | |
| | [Phishing Email Screenshots (25,000 screens)](https://huggingface.co/datasets/useKled/scam-emails) | Scam emails from inboxes in 11 Southeast Asian countries, screenshotted with the sender address and full body visible. | 25,000+ screenshots | |
| | [Booking Flow Screenshots (5,000 flows)](https://huggingface.co/datasets/useKled/searching-and-booking-step-by-step) | Ordered screenshot sequences of one flight, hotel or restaurant booking from search to review, each step described. | 5,000+ flows | |
| | [Shelf-Edge Price Tags (25,000 photos)](https://huggingface.co/datasets/useKled/shop-price-tags) | Close, straight-on photos of price labels and promo signs in shops across 11 Southeast Asian countries, digits sharp. | 25,000+ photos | |
| | [Slow Orbit Video, One Object (5,000 clips)](https://huggingface.co/datasets/useKled/slow-circle-video-around-one-object) | Continuous 30 to 60 second clips circling slowly around one ordinary object in even light, object centered throughout. | 5,000+ clips | |
| | [Staged Partial Occlusion (10,000 photos)](https://huggingface.co/datasets/useKled/something-partly-hidden) | Photos where one everyday object partly hides another, labeled with what is hidden and roughly how much is covered. | 10,000+ photos | |
| | [Retail Shelf Photos (50,000 photos)](https://huggingface.co/datasets/useKled/store-shelves) | Wide photos of stocked shelves in shops across 11 Southeast Asian countries, whole shelf in frame, prices readable. | 50,000+ photos | |
| | [SEA Street and Shop Signs (25,000 photos)](https://huggingface.co/datasets/useKled/street-and-shop-signs) | Road signs, shop fronts and warning signs across 11 Southeast Asian countries, every word legible, mixed scripts. | 25,000+ photos | |
| | [Suspected AI Profile Photos (10,000 screens)](https://huggingface.co/datasets/useKled/suspected-ai-profile-photos) | Dating and social profiles whose photo looks AI-generated, captured with the app interface, platform and tell noted. | 10,000+ screenshots | |
| | [Object Removal Sequences (10,000 sequences)](https://huggingface.co/datasets/useKled/take-one-thing-away-then-shoot-again) | Five-photo sequences from a propped, unmoving phone: exactly one object removed between shots, each removal named. | 10,000+ sequences | |
| | [Tampered Packaging Sets](https://huggingface.co/datasets/useKled/tampered-packaging) | Three-photo sets of packages that look opened or resealed: whole package, seal close-up, residue or misalignment. | volume on request | |
| | [SEA Transit Departure Boards (5,000 photos)](https://huggingface.co/datasets/useKled/transit-departure-boards) | Sharp photos of electronic departure boards and timetables at bus, rail and ferry stops and airports in Southeast Asia. | 5,000+ photos | |
| | [Consented Camera-Roll Corpus (200M files)](https://huggingface.co/datasets/useKled/upload-anything) | Photos and videos chosen by contributors from their own camera rolls, under a commercial training license. Screened. | 200,000,000+ files | |
| | [Vehicle Damage Sets (5,000 sets)](https://huggingface.co/datasets/useKled/vehicle-damage) | Damage photo sets for cars and motorbikes in Southeast Asia, whole vehicle to panel to close-up, with questionnaire. | 5,000+ photo sets | |
| | [Eight-View Object Orbits (10,000 objects)](https://huggingface.co/datasets/useKled/walk-around-one-object) | Eight-photo walks around one object on a table in even light, a frame every eighth turn, object and light unchanged. | 10,000+ objects | |
| | [Rain, Fog, Flood and Haze (10,000 photos)](https://huggingface.co/datasets/useKled/weather-degraded-scenes) | Streets, markets and roads photographed through monsoon rain, fog, flooding, heat haze and glare in Southeast Asia. | 10,000+ photos | |
| | [Entrances from the Street (10,000 files)](https://huggingface.co/datasets/useKled/building-entrances-and-access-points) | Photos and short clips of building entrances from the street: doors, entryways, signage and access points. | 10,000+ photos and clips | |
| | [New York City Potholes](https://huggingface.co/datasets/useKled/capture-nyc-street-potholes) | In-app photos of potholes on New York City streets, location-verified, some with a second shot of the street. | volume on request | |
| | [Finished Haircuts, Hair Only (50,000 files)](https://huggingface.co/datasets/useKled/capture-your-cut) | Completed haircuts and styling work framed on the hair, faces avoided, as photos and short clips across hair types. | 50,000+ photos and clips | |
| | [Delivery Drop-Off Spots (10,000 files)](https://huggingface.co/datasets/useKled/delivery-drop-off-locations) | Photos and short clips of where deliveries get left: lobbies, doorsteps and mailrooms, faces avoided. | 10,000+ photos and clips | |
| | [Consumer Drone City Video (100,000 files)](https://huggingface.co/datasets/useKled/map-your-flyover) | Drone video and aerial stills over streets, buildings and infrastructure, uploaded by pilots from their own flights. | 100,000+ clips and stills | |
| | [Pedestrian Decision Points (10,000 clips)](https://huggingface.co/datasets/useKled/navigation-decision-points) | Short clips of where a walker must choose: sidewalk forks, entrances, stairways and crossings, approach and choice. | 10,000+ clips | |
| | [Curbside and Street Parking (25,000 photos)](https://huggingface.co/datasets/useKled/parking-and-curbside-conditions) | Real curbside conditions: parked vehicles, loading zones, street parking and signage, as found in cities worldwide. | 25,000+ photos | |
| | [Consumer Dashcam Video (1.2M clips)](https://huggingface.co/datasets/useKled/upload-your-dashcam) | Dashcam video from contributors' regular drives, with stills, under a commercial training license. Real routes. | 1,200,000+ clips | |
| | [Pedestrian Walking Video (50,000 clips)](https://huggingface.co/datasets/useKled/urban-walking-navigation) | First-person clips of walking through sidewalks, crossings and dense streets, phone held forward at chest height. | 50,000+ clips | |
|
|
| ### Tools in Use (24) |
|
|
| Egocentric and third-person video of people doing ordinary tasks with their hands: cooking, folding, repairing, sorting, cleaning. Built for robot learning and world models. |
|
|
| | Dataset | What it is | Volume | |
| |---|---|---| |
| | [Tools Mid-Task (10,000 photos)](https://huggingface.co/datasets/useKled/a-tool-being-used) | Stills taken while a tool is in use, with tool, hands and the material being worked all visible, faces out of frame. | 10,000+ photos | |
| | [SEA Home Cooking, One Take](https://huggingface.co/datasets/useKled/cook-a-dish-from-start-to-finish) | Unbroken 5 to 10 minute videos of Southeast Asian home cooks making a dish they eat, raw ingredients to plated food. | volume on request | |
| | [Whole Shop Transactions on Video](https://huggingface.co/datasets/useKled/film-a-whole-shop-transaction) | Counter-view video of real sales in small Southeast Asian shops and stalls, from request to change, consented. | volume on request | |
| | [Verified Tradespeople at Work](https://huggingface.co/datasets/useKled/film-yourself-doing-your-actual-job) | Unbroken 5 to 10 minute videos of verified skilled workers doing a real paid job, hands in shot, explaining as they go. | volume on request | |
| | [Two-Handed Repair Video (5,000 clips)](https://huggingface.co/datasets/useKled/fix-something-with-tools-hands-in-shot) | Single-take 4 to 8 minute videos of repairs with hand tools, both hands and the workpiece in frame, no edits. | 5,000+ clips | |
| | [Narrated Task Video (5,000 clips)](https://huggingface.co/datasets/useKled/narrate-what-you-are-doing-while-you-do-it) | Unbroken 3 to 5 minute videos of everyday tasks, hands in frame, the person saying what they are doing as they do it. | 5,000+ clips | |
| | [Task Screen Recordings (10,000 recordings)](https://huggingface.co/datasets/useKled/screen-record-a-task-in-an-app) | Screen recordings of real tasks in contributors' own apps (order food, pay a bill, book a ride), tap trace attached. | 10,000+ recordings | |
| | [Failure and Recovery Video](https://huggingface.co/datasets/useKled/when-it-goes-wrong-film-the-fix) | Unbroken 2 to 4 minute videos that start when something goes wrong and follow what failed and what worked, narrated. | volume on request | |
| | [Vacuuming and Mopping Video (30,000 clips)](https://huggingface.co/datasets/useKled/capture-your-chores) | Everyday cleaning (vacuuming, mopping, wiping) filmed in real homes with the activity and the room in frame, faces out. | 30,000+ clips | |
| | [Counter Wiping Video (25,000 clips)](https://huggingface.co/datasets/useKled/clean-kitchen-surfaces) | People clearing a kitchen counter, wiping it down and putting things back, in good light, faces out of frame. | 25,000+ clips | |
| | [Home Upkeep Video (50,000 clips)](https://huggingface.co/datasets/useKled/document-your-routine) | Steady clips of routine household upkeep (cleaning, tidying, small maintenance jobs) in real homes, faces out of frame. | 50,000+ clips | |
| | [Repairs Start to Finish (25,000 clips)](https://huggingface.co/datasets/useKled/film-your-fix) | Complete repairs and installations in and around the home, camera on tools, materials and the work area. Real outcomes. | 25,000+ clips | |
| | [Garment Folding Video (100,000 clips)](https://huggingface.co/datasets/useKled/fold-laundry) | People folding real laundry, phone on the folding surface, faces out of frame. Deformable-object manipulation data. | 100,000+ clips | |
| | [Pick-and-Place Video (100,000 clips)](https://huggingface.co/datasets/useKled/object-transfer-between-surfaces) | Hands picking up everyday objects and setting them down on tables, shelves and counters in real homes. Raw video. | 100,000+ clips | |
| | [Articulated Object Video (100,000 clips)](https://huggingface.co/datasets/useKled/open-and-close-household-objects) | Hands opening and closing doors, cabinets, drawers and appliances in real homes. Raw demonstration video, third-person. | 100,000+ clips | |
| | [Drawer Reorganization Video (50,000 files)](https://huggingface.co/datasets/useKled/organize-a-drawer) | People emptying a drawer, sorting the contents, cleaning it and putting everything back in order, faces out of frame. | 50,000+ clips and stills | |
| | [Food Prep and Knife Work (30,000 clips)](https://huggingface.co/datasets/useKled/record-your-prep) | Food preparation filmed on the cutting board: chopping, peeling and slicing with hands in frame, faces out of frame. | 30,000+ clips | |
| | [Home Cooking Video (25,000 clips)](https://huggingface.co/datasets/useKled/record-your-recipe) | Unedited clips of home cooking from ingredients to the finished dish, filmed in contributors' own kitchens. | 25,000+ clips | |
| | [Book Sorting and Shelving (10,000 clips)](https://huggingface.co/datasets/useKled/shelve-your-stack) | Steady clips of people selecting, categorizing and placing books on shelves, with stills of the finished shelves. | 10,000+ clips | |
| | [Household Sorting Video (25,000 clips)](https://huggingface.co/datasets/useKled/sort-your-stuff) | People sorting clothes, tools and pantry goods into categories at home, sorting process in frame, faces out. | 25,000+ clips | |
| | [Trash Routine Video (50,000 clips)](https://huggingface.co/datasets/useKled/take-out-the-trash) | Continuous clips of the full trash routine: lift the bag out, tie it, fit a new liner, carry it out. Faces out of frame. | 50,000+ clips | |
| | [Household Tool Use Video (50,000 clips)](https://huggingface.co/datasets/useKled/tool-usage-non-industrial) | People using basic hand tools at home (scissors, screwdrivers, brooms), filmed on hands and tool, faces out of frame. | 50,000+ clips | |
| | [Grocery Unpacking Video (25,000 clips)](https://huggingface.co/datasets/useKled/unload-groceries) | People unpacking grocery bags and putting items away in fridges, cupboards and pantries, faces out of frame. | 25,000+ clips | |
| | [Dishwashing Video (75,000 clips)](https://huggingface.co/datasets/useKled/wash-the-dishes) | Complete dishwashing routines at the sink: rinse, soap, scrub, rinse, dry or load. Faces out of frame, real kitchens. | 75,000+ clips | |
|
|
| ### Financial & Legal Documents (14) |
|
|
| Native PDFs and photographed paper from real finances and real contracts, contributed by their owners: statements, invoices, payslips, leases, receipts, tax forms. |
|
|
| | Dataset | What it is | Volume | |
| |---|---|---| |
| | [Bank Statements as Native PDFs](https://huggingface.co/datasets/useKled/bank-statements) | Complete bank and card statements from contributors' own banking apps in 36 Western countries, every page, native PDF. | volume on request | |
| | [Business Invoices as Native PDFs](https://huggingface.co/datasets/useKled/business-invoices) | Real invoices sent or received at work, uploaded as original files by verified business users in 36 Western countries. | volume on request | |
| | [Forged Document Pairs](https://huggingface.co/datasets/useKled/document-forgery-pairs) | Pairs of a suspected forgery and a genuine document of the same class, altered regions boxed. No ID documents. | volume on request | |
| | [Fake Payment Proof Pairs](https://huggingface.co/datasets/useKled/fake-payment-confirmation-screenshots) | Pairs of a fake payment screenshot sent as proof of payment and a real confirmation from the same app, giveaway boxed. | volume on request | |
| | [Completed Paper Forms](https://huggingface.co/datasets/useKled/filled-in-paper-forms) | Full-page photos of paper forms filled in by hand (delivery slips, warranty cards, order pads), printed labels intact. | volume on request | |
| | [Business Email Compromise Attempts](https://huggingface.co/datasets/useKled/payment-redirect-scams) | Work messages asking the recipient to pay a new account or change bank details, sender and fake invoice included. | volume on request | |
| | [Payslips as Native Payroll PDFs](https://huggingface.co/datasets/useKled/payslips) | Real payslips uploaded as original payroll files by employees in 36 Western countries, earnings and deductions complete. | volume on request | |
| | [Signed Leases with Clause Marks](https://huggingface.co/datasets/useKled/rental-agreements) | Every page of signed rental and lease agreements from 36 Western countries, main clauses marked, tenant summary added. | volume on request | |
| | [Itemized Restaurant Bills (25,000 photos)](https://huggingface.co/datasets/useKled/restaurant-bills) | Flat overhead photos of itemized hospitality bills showing every line and the total, not the card slip. Many currencies. | 25,000+ photos | |
| | [Paper Retail Receipts (100,000 receipts)](https://huggingface.co/datasets/useKled/retail-receipts) | Overhead photos of contributors' own paper receipts, top to bottom in frame, faded and crumpled strips included. | 100,000+ receipts | |
| | [Real Spreadsheets People Built](https://huggingface.co/datasets/useKled/spreadsheets-you-made) | Spreadsheets people in 36 Western countries built for themselves, uploaded as is, structure intact, tagged by type. | volume on request | |
| | [Altered Receipt Pairs, Region Boxed](https://huggingface.co/datasets/useKled/suspected-fake-or-altered-receipts) | Pairs of a receipt the contributor believes was altered and a genuine one from a similar merchant, suspect region boxed. | volume on request | |
| | [Completed Tax Forms, Native Files](https://huggingface.co/datasets/useKled/tax-forms) | Completed tax forms as original files from contributors in 36 Western countries, every page and numbered box readable. | volume on request | |
| | [Utility and Telecom Bill PDFs (10,000 bills)](https://huggingface.co/datasets/useKled/utility-bills) | Electricity, water, internet and phone bills as original provider files from around the world, charge tables complete. | 10,000+ bills | |
|
|
| ### Everyday Writing (13) |
|
|
| Text people wrote for their own reasons, contributed by the author: emails, notes, reviews, recipes, how-to guides, support threads, code with rationale, code-switched messages. |
|
|
| | Dataset | What it is | Volume | |
| |---|---|---| |
| | [Home Recipes, Cook's Words (10,000 recipes)](https://huggingface.co/datasets/useKled/a-recipe-you-actually-cook) | Recipes written from memory by the people who cook them, with real amounts and timings and the step beginners get wrong. | 10,000+ recipes | |
| | [Code with Author Rationale (25,000 samples)](https://huggingface.co/datasets/useKled/code-with-your-explanation) | Author-contributed scripts, functions, queries and formulas, each with what it does, why, and the approach rejected. | 25,000+ samples | |
| | [Code-Switched Messages (50,000 pieces)](https://huggingface.co/datasets/useKled/code-switched-writing) | Informal messages written the way bilingual people text, switching languages mid-sentence, language pair declared. | 50,000+ pieces | |
| | [Draft and Final Writing Pairs (5,000 pairs)](https://huggingface.co/datasets/useKled/correct-this-text) | An early draft and the writer's own improved version of the same piece, with their explanation of what changed and why. | 5,000+ pairs | |
| | [Full Support Conversations (5,000 threads)](https://huggingface.co/datasets/useKled/customer-support-conversations) | Full support chats and email threads, both sides from opening to close, each tagged by the customer with how it ended. | 5,000+ conversations | |
| | [Author-Contributed Email (25,000 emails)](https://huggingface.co/datasets/useKled/emails-you-have-sent) | Emails contributed by the people who wrote them, pasted exactly as sent and tagged with what each was for. | 25,000+ emails | |
| | [Original How-To Guides (5,000 guides)](https://huggingface.co/datasets/useKled/how-to-instructions-you-write) | Step-by-step guides of at least 150 words by people who have the skill, numbered steps and what usually goes wrong. | 5,000+ guides | |
| | [Raw Meeting and Class Notes (10,000 sets)](https://huggingface.co/datasets/useKled/notes-you-took) | Notes people took in meetings and classes, pasted exactly as written: abbreviations, half-sentences, arrows, shorthand. | 10,000+ sets | |
| | [Real User Questions (50,000 questions)](https://huggingface.co/datasets/useKled/real-questions-you-d-ask-an-ai) | Questions people needed answered that week, typed with the reason behind each and the criteria a good answer must meet. | 50,000+ questions | |
| | [Reasoned Original Reviews (10,000 reviews)](https://huggingface.co/datasets/useKled/reviews-you-have-written) | Reviews of products and services the writer used, at least 80 words with the reasoning behind the opinion, typed in-app. | 10,000+ reviews | |
| | [Verbatim Scam Messages (25,000 messages)](https://huggingface.co/datasets/useKled/scam-messages-you-received) | Scam messages pasted exactly as they arrived on phones in 11 Southeast Asian countries, as plain text. The string. | 25,000+ messages | |
| | [Permissioned Workplace Documents](https://huggingface.co/datasets/useKled/workplace-documents) | Memos, reports and plans shared with permission by verified business users in 36 Western countries, tagged by type. | volume on request | |
| | [Native SEA-Language Text (25,000 pieces)](https://huggingface.co/datasets/useKled/writing-in-your-own-language) | Opinions, stories and explanations written by native or fluent speakers in their own languages, tagged, screened for MT. | 25,000+ pieces | |
|
|
| ### Research & Education (11) |
|
|
| Handwriting at scale, sketches, whiteboards, field notes, technical drawings and online pen strokes, for OCR, math and reasoning work. |
|
|
| | Dataset | What it is | Volume | |
| |---|---|---| |
| | [Charts Printed on Paper (5,000 photos)](https://huggingface.co/datasets/useKled/charts-printed-on-paper) | Straight-on phone photos of charts in newspapers, reports and posters, axes and legends readable, real paper and light. | 5,000+ photos | |
| | [Sketches Drawn to a Brief (5,000 sketches)](https://huggingface.co/datasets/useKled/hand-drawn-sketches) | Overhead photos of sketches drawn by hand to a prompt (floor plan, route, app screen, process), each described. | 5,000+ sketches | |
| | [Natural Handwriting Pages (500,000 pages)](https://huggingface.co/datasets/useKled/pages-of-your-handwriting) | Overhead photos of whole pages people wrote for themselves, in their own languages and scripts. Messy hands welcome. | 500,000+ pages | |
| | [Verified-Professional Explainers](https://huggingface.co/datasets/useKled/professional-domain-explainers) | Typed explanations of at least 300 words from verified professionals on what only doing the job teaches. Paste disabled. | volume on request | |
| | [Shared Technical Drawings (10,000 drawings)](https://huggingface.co/datasets/useKled/technical-drawings) | Photos and scans of technical drawings contributors may share (wiring, floor plans, assemblies), legends legible. | 10,000+ drawings | |
| | [Whiteboards After Meetings (10,000 photos)](https://huggingface.co/datasets/useKled/whiteboards) | Square-on, glare-free photos of whole whiteboards and flipcharts after lessons and meetings, handwriting and diagrams. | 10,000+ photos | |
| | [Online Handwriting Strokes (250,000 words)](https://huggingface.co/datasets/useKled/write-this-with-your-finger) | Words handwritten on a phone screen with a finger or stylus, captured as time-stamped stroke coordinates, not images. | 250,000+ words | |
| | [Note-Taking in Progress (150,000 files)](https://huggingface.co/datasets/useKled/capture-your-notes) | Students taking notes during lectures and study sessions, camera on the notebook or screen, as video and stills. | 150,000+ clips and stills | |
| | [Handwritten Notes and Math (750,000 pages)](https://huggingface.co/datasets/useKled/handwritten-notes-and-problem-solving) | Photographed and scanned pages of handwritten notes, worked math and diagrams from contributors worldwide. | 750,000+ pages | |
| | [Handwritten Field Notes (100,000 pages)](https://huggingface.co/datasets/useKled/log-your-fieldwork) | Well-lit, uncropped photos of handwritten field notes and data logs, page by page: tables, measurements, sketches. | 100,000+ pages | |
| | [Verified Student Coursework](https://huggingface.co/datasets/useKled/upload-your-homework) | Homework, essays and reports from contributors who verified student status with an institutional email, as images. | volume on request | |
|
|
| ### Speech & Audio (9) |
|
|
| Phone-recorded speech from a global base: read phrases, spoken digits and addresses, accents, code-switching, low-resource Southeast Asian languages and speaker-separated conversations. |
|
|
| | Dataset | What it is | Volume | |
| |---|---|---| |
| | [Korean Conversations (500 hours)](https://huggingface.co/datasets/useKled/conversations-in-korean) | Natural Korean conversations, 3 to 60 minutes, every participant consenting in-app, speaker turns marked and named. | 500+ hours | |
| | [Read Speech, 2,800 Phrases (500,000 clips)](https://huggingface.co/datasets/useKled/read-these-phrases-aloud) | Contributors reading on-screen phrases from a 2,800-item bank, one phrase per clip, mapped to its prompt. Many accents. | 500,000+ clips | |
| | [Consented Group Conversations](https://huggingface.co/datasets/useKled/record-a-conversation-with-someone) | Unscripted conversations of three minutes or more, every adult consenting in-app, speaker turns marked and named. | volume on request | |
| | [Speech in Real Noise (10,000 clips)](https://huggingface.co/datasets/useKled/record-in-a-noisy-place) | Contributors reading on-screen phrases in loud places (streets, cafes, markets, stations), with prompt and noise note. | 10,000+ clips | |
| | [Spoken Digits and Addresses (300,000 clips)](https://huggingface.co/datasets/useKled/say-numbers-names-and-addresses) | Prompted numbers, letter spellings, names and addresses read aloud, one item per clip, each mapped to a 3,425-item bank. | 300,000+ clips | |
| | [Low-Resource SEA Speech (50,000 clips)](https://huggingface.co/datasets/useKled/speak-in-a-less-common-language) | Native or fluent speech in regional languages of Indonesia, the Philippines and neighbors, prompted and free-form. | 50,000+ clips | |
| | [US Regional Accent Speech (25,000 clips)](https://huggingface.co/datasets/useKled/speak-in-your-regional-accent) | US contributors reading prompted phrases exactly as they speak at home, each clip tagged with where they are from. | 25,000+ clips | |
| | [Code-Switched Speech (10,000 clips)](https://huggingface.co/datasets/useKled/speak-mixing-two-languages) | Unscripted speech from bilingual Southeast Asians who switch languages mid-sentence, 30 seconds and up, pair declared. | 10,000+ clips | |
| | [Unscripted English Monologues (10,000 clips)](https://huggingface.co/datasets/useKled/talk-about-something-for-a-minute) | Off-the-cuff English speech of at least a minute on an on-screen topic, pauses and false starts left in, 5,000 topics. | 10,000+ clips | |
|
|
| ### Healthcare & Clinical (2) |
|
|
| De-identified intake forms and imaging studies, scoped with the buyer and reviewed before release. |
|
|
| | Dataset | What it is | Volume | |
| |---|---|---| |
| | [De-Identified Intake Forms](https://huggingface.co/datasets/useKled/medical-intake-and-forms-de-identified) | Images of medical intake forms, patient questionnaires and clinical assessments, identifiers removed before upload. | volume on request | |
| | [De-Identified Imaging Studies](https://huggingface.co/datasets/useKled/radiology-imaging-upload) | De-identified CT, MRI and ultrasound studies from US clinical-email-verified contributors, radiology report attached. | volume on request | |
|
|
| ## Links |
|
|
| - Catalog: [kled.ai/datasets](https://www.kled.ai/datasets?utm_source=huggingface&utm_medium=org_card) |
| - Data team: [team@kled.ai](mailto:team@kled.ai) |
| - Trust center: [trust.kled.ai](https://trust.kled.ai/) |
| - Terms: [kled.ai/terms-of-service](https://www.kled.ai/terms-of-service) |
| - X: [@useKled](https://x.com/useKled) |
|
|
