Published by Video Runtime · Source check 2026-09-21 · Evidence method · Suggest a correction
EDITORIALLY VERIFIED PRODUCT INFORMATION
About GIM-World
Long-horizon video world model rollouts that stay geometrically consistent through a fixed-size geometry-aware implicit memory. GIM-World compresses an arbitrarily long observation history into a fixed-size implicit memory supervised to encode 3D scene geometry, then conditions long autoregressive camera rollouts on that memory.
Key features
Code and first-person/third-person MIND checkpoints were released in September 2026. The memory design targets geometric and visual consistency where frame-retrieval and geometry-agnostic baselines drift.
Getting started
Install the documented environment, download the MIND checkpoints from ModelScope, then run the quick-start inference with the provided camera-trajectory format.
Real-time interaction
GIM-World compresses an arbitrarily long observation history into a fixed-size implicit memory supervised to encode 3D scene geometry, then conditions long autoregressive camera rollouts on that memory.
Pricing
Apache-2.0 code; self-hosted compute costs extra Repository code is published under Apache-2.0. Confirm separate model, dataset and dependency terms before reuse. No hosted price is inferred. Budget for the documented GPU, storage and setup requirements.
Limitations
Rollout quality still depends on the trained memory checkpoint and camera trajectory inputs; the README documents research rather than production deployment.
EDITORIAL VIEW
Editor's Verdict
GIM-World is included as a source-available implementation relevant to real-time generative video or interactive world systems.
Run the supplied example on representative hardware and review every linked licence before production use.
How It Works
- GIM-World compresses an arbitrarily long observation history into a fixed-size implicit memory supervised to encode 3D scene geometry, then conditions long autoregressive camera rollouts on that memory.
What We Like
- Code and first-person/third-person MIND checkpoints were released in September 2026.
- The memory design targets geometric and visual consistency where frame-retrieval and geometry-agnostic baselines drift.
Current Limitations
- Rollout quality still depends on the trained memory checkpoint and camera trajectory inputs; the README documents research rather than production deployment.
Pricing & Access
- Repository code is published under Apache-2.0. Confirm separate model, dataset and dependency terms before reuse.
- No hosted price is inferred. Budget for the documented GPU, storage and setup requirements.
Verification Summary
- Official repository and licence reviewed on 2026-09-21. Video Runtime has not installed or benchmarked this project.
- Performance and compatibility statements below are attributed to the maintainers and are not site measurements.
Technical notes & official performance
Official Performance Data
- The README states long-horizon rollouts stay geometrically coherent where explicit-memory baselines drift, but publishes no FPS or latency figure in the inspected sections. Video Runtime has not benchmarked it.
Technical Notes
- GIM-World is accepted to SIGGRAPH Asia 2026; the paper is on arXiv (2606.02436) and the work comes from the Kling Team at Kuaishou with Nanjing University and Tsinghua University.
Sources
- Official GitHub repository ↗github · accessed 2026-09-21
- Code licence ↗github · accessed 2026-09-21
- Official project page ↗official-docs · accessed 2026-09-21
Access, stage & mechanism
These are separate source-checked facts, not a live uptime monitor. Unverified fields are left open; transport and persistent sessions do not establish how frames are generated.
- Release stage
- Not verified
- Access mode
- Not verified
- Source availability
- Not verified
- Transport
- Not verified
- Interaction
- Not verified
- Session mode
- Not verified
- Generation mechanism
- Not verified
- Continuity method
- Not verified
- Licence status
- Not verified
- Tasks
- Not classified
CONTINUE EXPLORING
Related projects & models
Selected by research links and category, not by popularity.
Build long-horizon playable video worlds with camera control, prompt switching and explicit spatial memory.
Another frame streaming option in this section.Public repository; self-hosted setup and compatible GPU required
Train playable action-conditioned video world models with an open progressive causal recipe.
Another frame streaming option in this section.Public repository; self-hosted setup and compatible GPU required
Learn and reproduce the full pipeline for converting video diffusion backbones into interactive world models.
Another frame streaming option in this section.Public repository; self-hosted setup and compatible GPU required
Run an action-conditioned interactive world model with open-ended rollout on a single desktop GPU.
Another frame streaming option in this section.Public repository; self-hosted installation and compatible GPU requiredRead how official-source and hands-on records stay separate in our Methodology.