Step into any era. Change what happens next.
A photo, a letter, a map or a textbook page becomes a world you can enter. You stand where the picture was taken, the year is set, and the next decision is yours. History reacts — while every scene stays tied to its sources.
Runs in the browser · Polish and English narration · built in Gdynia, Poland
From a blank prompt to a first-person scene with real references, a narrator and a decision on screen. Recorded on the alpha build.
Sound on. Scenes, narration and images in the video were generated by the engine from a single prompt.
Finished paths can be generated in advance and reviewed by their authors. Visitors replay saved scenes; a new language may need an initial translation or narration pass. Creators use their own API keys. Institutions get editing, provenance and publication controls.
For visitors, students and the curious. No account, no key, no waiting.
For streamers, teachers, writers and friends on a Saturday night.
For museums, schools, publishers and cultural institutions.
The player sees a synthesis screen with accepted and rejected references, a live system log and fun facts about the place and time — so the wait itself teaches something.
The prompt is resolved into a time window and a point on the globe.
A first-person visual anchor: who you are, what you wear, what you hold.
Historically grounded people you will meet — profiled before the first scene.
Street-level imagery and reference photos are fetched and verified one by one.
A first-person scene with choices and mission goals as checkpoints.
4K scene images, narration and ambient sound for the era.
Sun position, weather, time zone and the era's atmosphere applied to the scene.
Synthesis view: references are accepted or rejected in the open, with a log you can read.
Define the endings you want, the number of choices per step and a budget. The autonomous builder runs on the server and fills the whole decision tree — typically within minutes — with a main path, side branches and a controlled number of dead ends. Close the browser; it keeps going and resumes after a restart.
Every branch remembers who made it. A published mission can accept new branches from other creators — with the author's authorisation and a notification.
Built in five months of nightly builds — over 680 production deployments so far.
Seven major eras and thirty-plus sub-eras. Click any point on Earth to get the place name, local time, time zone and weather for that moment, then travel there.
Every step is scored against what actually happened. Order a bust of Lenin in a Minsk office and the chronometer drops toward zero — you are told, in the open, that you are now in fiction. Two modes: historical keeps you on the rails, speculative lets you go off them knowingly.
The story is a graph, not a script. In Explorer you can open the tree and jump to any node of a finished path; in Creator you decide where it grows next.
Scenes cite where their facts and pictures came from. The engine queries encyclopaedias, museum and archive APIs and open image collections, prefers permissive licences, and stores the origin of every reference so an institution can contact the rights holder or swap the asset.
Select an object in the scene, give an instruction, re-render. Segmentation runs on our own server (no GPU needed), edits keep a history you can undo, and reference images can be attached without a limit.
A director model writes the shot list, scenes are animated, a narrator reads the hook and music is mixed underneath. The link opens full-screen, on phones too, and ends with a button into the game.
Missions are translated on the fly and cached per language, so the second visitor gets it instantly. Narration is generated per language with matching voices; subtitles follow.
On mobile, the phone's sensors move your head inside the scene; swipes navigate; touch-look is the fallback. On desktop the interface scale is a setting, so text is readable on a 13-inch laptop as well as a wall display.
A mission can leave the ground. Each scene has a realm — Earth, orbit, a planet or a place that never existed — and an algorithmic ephemeris positions the Sun, Moon and planets for the chosen date.
Reference images can be screenshots, scans or phone photos of physical pages. The engine reads them and builds the lesson as a playable story — ancient Mesopotamia from a first-year textbook, in first person, with quiz-like decisions.
The alpha has private, unlisted and public missions, source audits and approval of contributions to official missions. Player age guides generation; it is not an independent safety certification. Automated pre-publication age checks and a documented likeness-consent workflow are planned safeguards.
While a scene renders, the player reads and hears facts about the exact place and moment — preparing them for the decision ahead. In the "Now" era a geo-tagged news feed shows what is happening at the pin today.
Chronology panel: pin Warsaw, scrub to 1944, or to 2126.
Fun facts about the place and moment, read aloud while the scene loads.
A photo in a display case, a letter in a drawer, a caption on a wall. Turning one archive into a good interactive exhibit today takes a curator, a writer, a designer, a researcher, a developer and five tools. OmniHistory collapses that into one workflow — and keeps the curator in charge of facts, sources, rights and publication.
Photos of exhibits, letters, maps, textbook pages — taken with a phone on site if you like.
Narrative graph, scenes, characters, decisions and alternative paths — hundreds of possible passages, generated once.
Correct text and images, inspect source licences, choose mission visibility and review contributions before they reach visitors.
In a browser, on a phone or on a gallery display. A prepared path can be replayed without regenerating its saved scenes.
It is 1939. You stand where the photograph was taken. Follow events or try a different choice. History reacts. The source trail and consistency indicator help you examine where the reconstruction draws on evidence and where interpretation begins.
Tested with children: the fun was not watching the story — it was deciding what happens next. Waiting for the next scene only raised the tension.
Book a demo with your materials →"The most valuable asset of a museum is not the building. It is the stories hidden in thousands of photographs, letters, maps and documents." — from the OmniHistory stage pitch, September 2026
Generative models get better every month — and all of them can be plugged in here. The value is in what sits around them.
Scenes preserve reference sources and generation history where available. Curators can inspect and correct them; an official badge identifies the account, not a guarantee of accuracy.
Edit scenes, control visibility, approve contributions and publish deliberately — these are tools, not just instructions in a prompt.
Each institution that publishes makes Explorer more valuable for the next one. The content network compounds; a single model does not.
The engine integrates several text, image and video providers, with local options for selected tasks. Provider availability and capabilities differ.
The alpha is live at a private address. Institutional pilots decide what gets built first.
Write to us for early access, an institutional demo with your own materials, or a partnership conversation. We answer personally.
Or simply: lazniak@hexart.pl