Tesana claims Muranyi-3 makes text-to-playable 3D worlds 70% faster
Tesana says Muranyi-3 turns plain-text prompts into playable 3D worlds with generated logic, NPCs, reference-art conditioning, and collaboration links. The launch also claims 90% fewer common errors than the prior model, but the evidence here comes from a promotional thread.

TL;DR
- On its homepage, Tesana's Muranyi-3 product page says the model builds playable games from a single prompt, while hasantoxr's launch thread frames it as text-to-playable 3D world generation with characters, physics, UI, and logic.
- According to hasantoxr's systems post, Muranyi-3 is pitched as a system generator, not just a scene maker, with NPC behaviors, quest structures, collision detection, and other game logic generated from the prompt.
- Both Tesana's capabilities section and hasantoxr's refinement demo say creators can keep editing the same world in plain language, with changes applied live instead of rebuilding from scratch.
- Style control is a first-class feature in both Tesana's reference images section and hasantoxr's reference art demo, which show uploaded concept art steering characters, environments, and UI.
- The speed and quality numbers come from promotion, not an eval card: hasantoxr's metrics post claims 90% fewer common game generation errors and roughly 70% faster generation than the previous model, while Tesana's research page mainly explains the older Muranyi-2 architecture.
You can browse playable examples on Tesana's homepage, read the older Muranyi-2 research explainer, and check the pricing page, which already spells out export rights, Godot source downloads, and faster generations on higher tiers.
Playable worlds
Tesana's clearest official claim is on the homepage: Muranyi-3 is an end-to-end game engine that generates mechanics, environments, characters, UI, and systems from one prompt. That lines up almost word for word with hasantoxr's launch thread and hasantoxr's onboarding post, which show the pitch as type a concept, generate, then start playing.
The homepage breaks the stack into a few concrete pieces:
- Create worlds: prompts generate environments and physics, per Tesana's capabilities section
- NPCs and quests: plain-language requests become behaviors, quests, and logic, per Tesana's capabilities section and hasantoxr's systems post
- Reference images: uploaded art guides the visual direction, per Tesana's capabilities section
- Expand endlessly: mechanics, difficulty, visuals, and quests can be changed later, per Tesana's capabilities section
The odd bit is the sourcing. Tesana's public site is specific about features, but the 90% and 70% launch numbers in circulation are coming through hasantoxr's metrics post, not a public benchmark page or changelog entry.
Live iteration
The strongest creative workflow claim here is continuity. Instead of treating generation as a one-shot export, hasantoxr's refinement demo shows follow-up prompts like "make the enemies faster" or "add a snow biome," while Tesana's homepage says the live world updates instantly so creators can iterate without starting over.
That matters because it shifts Muranyi-3 closer to an editable engine surface than a prompt-to-demo toy. Tesana's own Muranyi-2 research page describes the series as code-based game models that translate descriptions into structured game code, which helps explain why the company keeps stressing systems and logic rather than image generation.
Style control
Muranyi-3 is also leaning hard on authorship. Tesana's homepage says creators can upload reference art or character concepts to steer the look of environments, characters, and UI, and hasantoxr's reference art demo uses that exact feature as the bridge from generic AI output to something that feels art-directed.
For designers, this is the most concrete differentiator in the launch materials. Text-only generation is common now. Reference-conditioned worldbuilding, where the same image steers the game's characters, environment, and interface, is the part Tesana keeps presenting as identity control rather than mere generation.
Sharing, exports, and pricing
The last useful reveal is that Tesana's business model already exposes some of the practical workflow. hasantoxr's collaboration post says worlds can be shared by link and edited collaboratively across devices, while the official pricing page says the $20 Pro plan includes public sharing, commercial use rights, export to macOS, Windows, and web builds, plus downloadable Godot source code for original games.
The higher $50 Max tier adds faster generations and custom 3D model, audio, and sprite imports, according to Tesana's pricing page. That is more operational detail than the launch thread gives, and it is the clearest sign that Tesana is selling Muranyi-3 as a publishable toolchain, not just a browser demo.