conv.

All stories
TechQuiet 2d · day 3

Tencent releases WorldClaw, agentic 3D open-world generation system

Tencent's Hunyuan3D team publishes WorldClaw, which uses LLMs and image models to generate large-scale 3D game worlds procedurally.

Tencent releases WorldClaw, agentic 3D open-world generation system
tencent-hunyuan.github.io

Conversation activity · last 3 days presspostscomments the crowd ↓

Peak 8 items in one hour at Aug 11, 6 PM; 24 items over 3 days (2 press · 10 posts · 12 comments) Aug 11, 7 AM — 1 item · 1 post — Hacker News 1Aug 11, 8 AM — no itemsAug 11, 9 AM — no itemsAug 11, 10 AM — no itemsAug 11, 11 AM — no itemsAug 11, 12 PM — no itemsAug 11, 1 PM — no itemsAug 11, 2 PM — no itemsAug 11, 3 PM — no itemsAug 11, 4 PM — no itemsAug 11, 5 PM — 3 items · 2 press · 1 post — Newswires 2, Hacker News 1Aug 11, 6 PM — 8 items · 4 posts · 4 comments — Hacker News 4, Mastodon 4Aug 11, 7 PM — 3 items · 3 comments — Hacker News 3Aug 11, 8 PM — no itemsAug 11, 9 PM — 1 item · 1 comment — Hacker News 1Aug 11, 10 PM — 1 item · 1 comment — Hacker News 1Aug 11, 11 PM — 1 item · 1 post — Mastodon 1Aug 12, 12 AM — no itemsAug 12, 1 AM — no itemsAug 12, 2 AM — no itemsAug 12, 3 AM — 2 items · 1 post · 1 comment — Hacker News 1, Mastodon 1Aug 12, 4 AM — no itemsAug 12, 5 AM — 1 item · 1 comment — Hacker News 1Aug 12, 6 AM — 1 item · 1 post — Mastodon 1Aug 12, 7 AM — 1 item · 1 comment — Hacker News 1Aug 12, 8 AM — 1 item · 1 post — Mastodon 1Aug 12, 9 AM — no itemsAug 12, 10 AM — no itemsAug 12, 11 AM — no itemsAug 12, 12 PM — no itemsAug 12, 1 PM — no itemsAug 12, 2 PM — no itemsAug 12, 3 PM — no itemsAug 12, 4 PM — no itemsAug 12, 5 PM — no itemsAug 12, 6 PM — no itemsAug 12, 7 PM — no itemsAug 12, 8 PM — no itemsAug 12, 9 PM — no itemsAug 12, 10 PM — no itemsAug 12, 11 PM — no itemsYesterday, 12 AM — no itemsYesterday, 1 AM — no itemsYesterday, 2 AM — no itemsYesterday, 3 AM — no itemsYesterday, 4 AM — no itemsYesterday, 5 AM — no itemsYesterday, 6 AM — no itemsYesterday, 7 AM — no itemsYesterday, 8 AM — no itemsYesterday, 9 AM — no itemsYesterday, 10 AM — no itemsYesterday, 11 AM — no itemsYesterday, 12 PM — no itemsYesterday, 1 PM — no itemsYesterday, 2 PM — no itemsYesterday, 3 PM — no itemsYesterday, 4 PM — no itemsYesterday, 5 PM — no itemsYesterday, 6 PM — no itemsYesterday, 7 PM — no itemsYesterday, 8 PM — no itemsYesterday, 9 PM — no itemsYesterday, 10 PM — no itemsYesterday, 11 PM — no itemsToday, 12 AM — no itemsToday, 1 AM — no itemsToday, 2 AM — no itemsToday, 3 AM — no itemsToday, 4 AM — no itemsToday, 5 AM — no itemsToday, 6 AM — no itemsToday, 7 AM — no itemsToday, 8 AM — no itemsToday, 9 AM — no itemsToday, 10 AM — no itemsToday, 11 AM — no itemsToday, 12 PM — no items 8 items · Aug 11 6 PM
Aug 12yesterdaytodaynow · 1:07 PM
Every item, in order 24

Latest coverage newest 4 of 4 items

What to know

  • WorldClaw combines image generation with 3D extraction (SAM3D) to procedurally compose and populate large-scale game worlds—a pipeline approach not commonly seen.
  • The system is a Python wrapper around external models with no public code; it does not introduce new foundational models.
  • Early reactions acknowledge technical merit but highlight practical limits: generated worlds show asset placement errors, cartoonish aesthetics, and lack the narrative depth of hand-crafted environments.
  • Commenters debate whether procedural generation can meaningfully assist indie developers or whether it primarily targets mass-production of lower-quality games (gacha style).

How it unfolded

  1. A commenter raises philosophical concerns about attribution when AI generates content: when you see a well-designed building or easter egg, was it made by a human or generated autonomously? This complicates the value proposition for indie developers.

    “When you see an elegantly designed building, environment or easter egg in the game, was that really made by a human (or even prompted by a human), or did the AI generate it autonomously as part of a much larger generation step?”

    2001zhaozhao · Hacker News ↗
  2. Multiple commenters identify flaws in the generated worlds: buildings placed on water, cartoonish terrain, and questions about whether outputs are cherry-picked examples. One asks whether the system can generate non-MMO cartoony styles.

    “In both the Autumn and Winter examples in the hero images, the algo seemed to have placed buildings on the water in the foreground”

    cobertos · Hacker News ↗
  3. A technical commenter clarifies that WorldClaw is not a model but Python scripts calling external models. The genuine novelty identified is using image models for composition, then extracting objects to 3D via SAM3D—an approach not commonly seen.

    “In case it's not obvious, this isn't a model, this is python scripts that call out to models (code not available)… an image model performs the composition (which image models are really good at), and then you extract the objects into 3d…”

    avaer · Hacker News ↗
  4. A Hacker News commenter argues that open worlds are strongest with hand-placed details and storytelling, citing Skyrim and Cyberpunk as examples superior to procedurally generated content like Starfield.

    “It looks impressive. However open worlds are at their best with hand placed details and environmental storytelling. Look at Skyrim/Cyberpunk vs. Starfield where most of the world is proc gen'd.”

    cautiouscat · Hacker News ↗
  5. A Mastodon user dismissed WorldClaw as overhyped, characterizing it as 'more polygons dressed up in buzzword salad' with no meaningful innovation.

    “Ah, yes, the earth-shattering revolution of "Agentic 3D # OpenWorld # Generation at Scale" — because what the world really needs is another # overhyped attempt to make computers spit out endless, mind-numbing virtual landscapes”

    ngate@mastodon.social · Mastodon ↗
  6. Tencent's Hunyuan3D team released WorldClaw, a system for agentic 3D open-world generation at scale. The project was posted to Hacker News and quickly gained traction.

What people are saying verbatim

“In case it's not obvious, this isn't a model, this is python scripts that call out to models (code not available).”

avaer, HN commenter · Hacker News ↗

“an image model performs the composition (which image models are really good at), and then you extract the objects into 3d via things like SAM3D before placing them in the world, which is pretty interesting.”

avaer, HN commenter · Hacker News ↗

“It looks impressive. However open worlds are at their best with hand placed details and environmental storytelling. Look at Skyrim/Cyberpunk vs. Starfield where most of the world is proc gen'd.”

cautiouscat, HN commenter · Hacker News ↗

“what the world really needs is another # overhyped attempt to make computers spit out endless, mind-numbing virtual landscapes 🌍🔨”

ngate@mastodon.social, Mastodon user · Mastodon ↗

“When you see an elegantly designed building, environment or easter egg in the game, was that really made by a human (or even prompted by a human), or did the AI generate it autonomously as part of a much larger generation step?”

2001zhaozhao, HN commenter · Hacker News ↗

“In both the Autumn and Winter examples in the hero images, the algo seemed to have placed buildings on the water in the foreground”

cobertos, HN commenter · Hacker News ↗

“I wonder if there can be more style applied than just the "MMO cartoony" style? Like medieval Witcher or realistic Cyberpunk buildings?”

gosolozero, HN commenter · Hacker News ↗

The conversation unedited, verbatim

86 comments claimed on the threads · 15 collected here from 2 sites · 15 voices below over time ↑

Hacker News 12 Peak 4 items in one 2h at Aug 11, 5 PM; 12 items over 3 days

Where people stand verbatim

The conversation centers on WorldClaw's actual technical contribution versus hype. Skeptics dismiss it as buzzword-laden, while others concede the image-to-3D extraction idea has merit but flag quality issues and fundamental limits of procedural worlds compared to hand-crafted games.

The dispute Whether procedurally generated worlds represent a meaningful advance for game development or a distraction that cannot replicate the quality and intentionality of hand-crafted environments.

many voices

The system's core innovation—using image models for composition and SAM3D for extraction—is genuine and worth attention, even if execution has flaws.

  • “an image model performs the composition (which image models are really good at), and then you extract the objects into 3d via things like SAM3D before placing them in the world, which is pretty interesting.”

    avaer · Hacker News ↗
  • “That's exactly how I've used tools like Tripo and Meshy… Use any image generation tool (or any image) and have a multimodal LLM like Gemini extract individual parts of the image”

    pinkmoonx · Hacker News ↗
many voices

Procedurally generated worlds cannot match hand-crafted games in detail and narrative design; WorldClaw's outputs are unimpressionably cartoonish and error-prone.

  • “It looks impressive. However open worlds are at their best with hand placed details and environmental storytelling. Look at Skyrim/Cyberpunk vs. Starfield where most of the world is proc gen'd.”

    cautiouscat · Hacker News ↗
  • “In both the Autumn and Winter examples in the hero images, the algo seemed to have placed buildings on the water in the foreground”

    cobertos · Hacker News ↗
some voices

This is hype-driven marketing with no real innovation—just connecting existing models and calling it revolutionary.

  • “Ah, yes, the earth-shattering revolution of "Agentic 3D # OpenWorld # Generation at Scale" — because what the world really needs is another # overhyped attempt to make computers spit out endless, mind-numbing virtual landscapes”

    ngate@mastodon.social · Mastodon ↗
  • “this isn't a model, this is python scripts that call out to models (code not available). It's mostly "did you know you can attach an LLM to a PCG system?"”

    avaer · Hacker News ↗
Still unanswered
  • Are the generated examples cherry-picked, or is asset placement and terrain quality consistently this flawed?
  • Can WorldClaw generate styles beyond 'MMO cartoony'—e.g., medieval, realistic, Cyberpunk aesthetics?
  • How much human effort is required post-generation to make these worlds playable, and does that undermine the time-savings pitch?
  • Aug 12
  • I just left the website open by mistake and went to take a walk. The temp got 80 degree celcius. My laptop's chasis got really hot

    shuvrojitHacker News2d agoview on Hacker News ↗
  • WorldClaw — Agentní generování 3D otevřeného světa ve velkém měřítku https:// tencent-hunyuan.github.io/Huny uan3D-WorldClaw/

    365tipu@mastodonczech.czMastodon · fosstodon.org2d agoview on Mastodon ↗
  • Again, that Claude web design aesthetic is just seemingly everywhere right now. It takes me out of what I'm reading as soon as you see the tells.

    ablationHacker News2d agoview on Hacker News ↗
  • LLMs are a PCG system. This is one of my pet peeves, I've been working on a procedural generation constraint system for quite a while (way before LLMs) as a hobby, that came with a domain specific language, where you could specify constraints.Things like the red key must be reachable before the red door, and there should be no path that bypasses…

    torginusHacker News2d agoview on Hacker News ↗
  • Aug 11
  • Text to Open world design. # deathoftheinternet # aislop # gamedesign # prompt https:// tencent-hunyuan.github.io/Huny uan3D-WorldClaw

    n_dimension@infosec.exchangeMastodon · toot.community2d agoview on Mastodon ↗
  • The water levels don't change seasonally..So I'm guessing it's a very high rainfall area? In which case erosion and nutrient leaching are bound to be issues for the agentic farmers. I wonder if they've thought to build soil with perennial ground cover species..But as I say this, I realise the tree species seem to change very dynamically from…

    thymine_dimerHacker News2d agoview on Hacker News ↗
  • > Still, think most people don't realize how good LLMs are at 3D these days, especially Fable 5Can you explain more? Can I use Fable 5 to generate a 3d model from a couple of photographs?

    doctobogganHacker News2d agoview on Hacker News ↗
  • This kind of content generation can be quite cool to enable new game ideas for indie developers that were only possible with AAA before. But it also makes it hard to gauge the amount of human work that went into the game. When you see an elegantly designed building, environment or easter egg in the game, was that really made by a human (or even…

    2001zhaozhaoHacker News2d agoview on Hacker News ↗
  • That’s exactly how I’ve used tools like Tripo and Meshy.Use any image generation tool (or any image) and have a multimodal LLM like Gemini extract individual parts of the image “isolate with a transparent background” - then you can use those images for image-to-3d in those tools.Don’t forget to use low or “smart” poly features otherwise you get…

    pinkmoonxHacker News2d agoposition 1 ↓view on Hacker News ↗
  • In both the Autumn and Winter examples in the hero images, the algo seemed to have placed buildings on the water in the foregroundIn the summer example in the hero images, the building placement + small pockets of water on the left looks odd and low attention to detail. A similar poor quality result as if an uncaring human used a scatter…

    cobertosHacker News2d agoposition 2 ↓view on Hacker News ↗
  • Interesting prompt to world pipeline. I wonder if there can be more style applied than just the "MMO cartoony" style? Like medieval Witcher or realistic Cyberpunk buildings?

    gosolozeroHacker News2d agoview on Hacker News ↗
  • In case it's not obvious, this isn't a model, this is python scripts that call out to models (code not available).It's mostly "did you know you can attach an LLM to a PCG system?", but there is one idea here you don't see much of: an image model performs the composition (which image models are really good at), and then you extract the objects into…

  • It looks impressive. However open worlds are at their best with hand placed details and environmental storytelling. Look at Skyrim/Cyberpunk vs. Starfield where most of the world is proc gen’d.I get that you run this, and then edit it. But the generated villages just aren’t interesting in my opinion. It’s probably great for tencent’s market where…

    cautiouscatHacker News2d agoposition 2 ↓view on Hacker News ↗
  • The assets look good, but for some reason the terrain looks too cartoonish to me.

  • Ah, yes, the earth-shattering revolution of "Agentic 3D # OpenWorld # Generation at Scale" — because what the world really needs is another # overhyped attempt to make computers spit out endless, mind-numbing virtual landscapes 🌍🔨. Spoiler alert: it’s just more polygons dressed up in buzzword salad 🥗🤖. https:// tencent-hunyuan.github.io/Huny…

    ngate@mastodon.socialMastodon · toot.community2d agoposition 3 ↓view on Mastodon ↗