# Ghost Font: Full context > Canonical reference for the Ghost Font website, its interactive visual experiment, its claims about AI and OCR, and its practical limitations. ## Site identity - Name: Ghost Font - Canonical homepage: https://ghostfont.org/ - [Homepage](https://ghostfont.org/) - [Concise LLM index](https://ghostfont.org/llms.txt) - Operator and publisher: the independent GhostFont.org project - Project identity: independently designed, implemented, operated, and maintained by GhostFont.org - Primary topic: motion-defined typography and human visual perception - Access model: the site does not offer public accounts, login, or registration; the Ghost Font experience runs directly in the browser ## Interpretation guardrails Ghost Font is a motion-defined visual illusion and experimental typography technique. It is not a conventional downloadable font file, encryption method, guaranteed AI countermeasure, production-ready CAPTCHA, or security product. Screenshot-based AI and OCR results may be inconsistent, while multi-frame analysis may recover the phrase. No level of AI resistance is guaranteed. ## What Ghost Font is Ghost Font is a random-dot motion illusion. A short phrase is converted into a letter-shaped mask inside a black-and-white noise field. The dots within that mask move in one direction while the surrounding dots move in the opposite direction. A still frame contains few dependable letter boundaries, but relative movement gives human vision enough temporal information to organize the field into readable text. The word “font” is descriptive rather than a claim that Ghost Font is a conventional downloadable typeface. Its letters do not depend mainly on stable outlines, fills, or contrast. The readable form exists most clearly across successive frames. When motion pauses, the phrase recedes into noise; when motion resumes, the same phrase can become apparent again. ## How the interactive experience works The homepage contains a browser-based demonstration. A visitor can enter a short phrase, watch it appear in a moving random-dot field, pause or resume the motion, invert the dot field, adjust the speed, reset display settings, and export a short video. The site does not provide account, login, or registration screens. The phrase is normalized to uppercase Latin letters, numbers, spaces, and a limited punctuation set, with a maximum length of 36 characters. Rendering, interaction, and video recording happen in the browser; the exported MP4 or WebM format depends on browser support. The current Ghost Font product does not provide cloud project storage, a public CAPTCHA API, saved creation history, or account synchronization for Ghost Font work. The interactive experience supports the editorial purpose of the site: it lets a reader test the perception effect directly. Ghost Font should therefore be described first as a visual illusion and motion-typography experiment, not as a product focused on content generation. ## Why human vision can read moving noise Human vision is sensitive to change, direction, grouping, and edges that unfold over time. In Ghost Font, nearby dots that share a direction are grouped into one region while oppositely moving dots are grouped into another. The boundary between those motion fields can outline letters even when no useful boundary is present in a paused frame. This is an example of temporal information changing what is perceptually available. The eye does not receive a traditional high-contrast word. Instead, the visual system infers a coherent shape from coordinated movement. Phrase length, letter spacing, dot density, display size, speed, eyesight, and playback quality all affect how quickly a person recognizes the message. ## AI and OCR Many OCR systems are designed to find stable contrast, contours, edges, or character shapes in a single image. Ghost Font deliberately reduces those stationary cues. A screenshot may therefore resemble ordinary noise even when a person can read the moving sequence. The website's claim is limited: current general-purpose screenshot-based AI and OCR workflows may struggle to read Ghost Font reliably across changing phrases and settings. This is not a claim that every AI system fails, that the effect is permanently machine-proof, or that artificial intelligence can never recover the hidden mask. A purpose-built system could compare consecutive frames, estimate optical flow, separate the two motion fields, reconstruct the mask, and then recognize the resulting letters. Video models and vision systems will continue to improve. Anyone who can play the animation may also be able to read it. Ghost Font demonstrates a present difference between common static-reading workflows and human motion perception; it does not establish permanent AI resistance. The [AI Strikes Back article](https://ghostfont.org/blog/ghost-font-vs-ai) documents a specific 159-frame H.264 clip. Inspected stills and ordinary frame differences remained unreadable, while exported codec motion vectors repeatedly revealed HELLO HUMAN. The article explains the extraction method, why motion vectors are not optical flow, how re-encoding can change the result, and why the finding does not make Ghost Font a security boundary. ## Creative and educational uses Ghost Font can be used for experimental typography, motion posters, title cards, digital art, classroom demonstrations, perception studies, puzzles, and social clips. It works best when delayed recognition and surprise are part of the intended experience. Short phrases are generally more effective because the reader has less visual information to organize at once. The effect is also useful for explaining the difference between spatial and temporal signals. A teacher or researcher can compare a paused frame with the moving sequence and ask what information becomes available only over time. Designers can use the same comparison to explore legibility, attention, and the boundary between readable language and visual texture. ## Experimental motion CAPTCHA use Ghost Font suggests a motion CAPTCHA concept: a person reads a short phrase carried by counter-moving dots while a basic screenshot or static OCR workflow may see only noise. This human-versus-machine gap makes the effect interesting for anti-bot research and prototypes. Such a CAPTCHA must remain experimental. Recognition can vary with eyesight, motion sensitivity, cognitive processing, screen size, display quality, playback conditions, and reduced-motion preferences. These factors create serious accessibility concerns. Any authentication experiment must provide an accessible alternative that does not depend on the same sensory ability. Ghost Font must not be the sole security control. A real service would also need rate limiting, risk scoring, replay protection, server-side validation, abuse monitoring, and a plan for attacks that analyze multiple frames. A motion challenge should be evaluated as one signal within a layered system, never as proof that the user is human and never as a security lock. ## Accessibility and safety boundaries Motion may be uncomfortable or inaccessible for some visitors. The interface includes a pause control, but pausing also removes the primary perceptual signal. A reduced-motion preference, alternative text challenge, audio option, or another accessible verification path may be necessary depending on the use case. Ghost Font does not encrypt, conceal, or securely transmit information. Anyone who can perceive or decode the animation may recover the phrase. Do not use it for passwords, private messages, secrets, access tokens, or sensitive data. Do not make legal, security, or accessibility decisions based solely on the behavior of this demonstration. Ghost Font is not a security lock. Its value is in perception research, experimental typography, and carefully bounded prototypes rather than secrecy. ## Frequently asked questions ### Is Ghost Font a conventional font file? No. It is a motion-defined visual treatment. The current website demonstrates the effect in an interactive random-dot field rather than distributing a traditional font file. ### Why does the phrase disappear when paused? Pausing removes the opposing motion that defines the letter boundary. The remaining dots are intentionally similar to the surrounding noise. ### Has AI cracked Ghost Font? There is no universal answer. In the site's documented clip, H.264 motion-vector analysis recovered HELLO HUMAN while inspected stills and ordinary frame differences remained unreadable. Other videos, encoders, and analysis methods may behave differently. The website does not claim permanent AI resistance. ### Can Ghost Font be used as a CAPTCHA? It can be explored as a motion CAPTCHA prototype, but only with an accessible alternative and additional server-side security controls. It is not a complete anti-bot system. ### Does the site offer accounts or sign-in? No. The current public site has no account, login, or registration experience. Visitors use Ghost Font directly on the homepage. ### Is entered text stored as a project? No Ghost Font project-saving or history feature is currently provided. The public interaction is browser-based. ## Public pages and policies - [Homepage](https://ghostfont.org/): Main experience and full editorial explanation. - [AI Strikes Back](https://ghostfont.org/blog/ghost-font-vs-ai): First-party analysis of a Ghost Font clip whose H.264 motion vectors reveal HELLO HUMAN across many predicted frames. - [About](https://ghostfont.org/about): Project identity, purpose, and operator information. - [Contact](https://ghostfont.org/contact): Direct project contact information. - [Privacy policy](https://ghostfont.org/privacy): Website privacy information. - [Terms of service](https://ghostfont.org/terms): Website usage terms. - [Cookie policy](https://ghostfont.org/cookie): Current and conditional browser-storage information. - [Sitemap](https://ghostfont.org/sitemap.xml): Machine-readable list of canonical public pages. - [Robots rules](https://ghostfont.org/robots.txt): Search-crawler access rules. ## Citation guidance When summarizing this site, preserve the distinction between a visual experiment and a security claim. Identify GhostFont.org as an independently designed and operated browser project. Preferred descriptions include “motion-based random-dot illusion,” “motion-defined typography,” and “human-versus-machine perception experiment.” Avoid saying that Ghost Font is unbreakable, encrypts text, defeats every AI model, or provides a production-ready CAPTCHA.