- Services
Code Video Model is a research system from Microsoft Research that turns a text prompt into an editable Three.js program, renders a low-poly proxy video in under a minute, then conditions the frozen MiniMax-H3 model on that proxy and a reference image to produce the final clip. It is training-free: MiniMax-H3 is not fine-tuned and no control module is trained. Output is 960x544 video with audio, and the code is on GitHub under Apache-2.0. Running it needs Ubuntu 22.04, an NVIDIA GPU with a CUDA 12.6 driver, the MiniMax-H3 checkpoint and an OpenRouter API key.
A workable dense video generator from Microsoft Research. Treat the modality benchmarks above as the leading indicator of fit — composite scoring across modalities is still maturing. Newly released, so production-readiness is still being shaken out.
Generated from this model’s benchmarks and ranking signals. Editor reviews refine it over time.
No benchmark data available for this model yet.
The top devices for this model at 4-bit, ranked by fit and speed.
| Device | Grade | VRAM |
|---|---|---|
| ACEMAGIC M1A Pro (i9-13900HK + ARC A770)ACEMAGIC | SS | 0.5 GB |
| Acer Veriton GN100 AI MiniAcer | SS | 0.5 GB |
| AMD Instinct MI300XAMD | SS | 0.5 GB |
| AMD Instinct MI325XAMD | SS | 0.5 GB |
| AMD Instinct MI355XAMD | SS | 0.5 GB |
Cheapest current cloud rentals with at least 1 GB VRAM, refreshed hourly.
Advertising disclosure: we earn commissions when you shop through the links below.
| Option | Cost / GPU-hour |
|---|---|
NVIDIA GeForce RTX 3060Vast.ai · Spot · 12 GB VRAM | $0.05 |
NVIDIA GeForce RTX 3060Vast.ai · On-Demand · 12 GB VRAM | $0.07 |
NVIDIA GeForce RTX 3060 TiVast.ai · Spot · 8 GB VRAM | $0.07 |
NVIDIA GeForce RTX 4060Vast.ai · Spot · 8 GB VRAM | $0.08 |
NVIDIA GeForce RTX 2080 TiVast.ai · Spot · 11 GB VRAM | $0.08 |
Per-GPU rate across RunPod, the Vast.ai marketplace, DigitalOcean, and Vultr.
Spot tier is interruptible. Plan for restarts when comparing against on-demand prices.
A coding agent writes each shot as a Three.js program that lays out the scene, animates objects, places the camera and computes dynamics from equations. Rendering the program yields a proxy video and a reference image. Early-step conditioning attends to the proxy only during the first 5 of MiniMax-H3's 20 denoising steps, so the final clip follows the proxy's structure while keeping the reference image's appearance. The team also released CodeVideoBench, a 22-case benchmark across gaming, scene editing, world-model data, physics grounding, bullet time and 3D/4D reconstruction.
Explore the Provider
Aggregate stats, leaderboard, release timeline, and benchmark coverage across every Microsoft Research model we track.
Check which of your devices can run this model and how fast.
Compare hosted per-token prices before you commit to local hardware.
Break-even math for self-hosting, renting a GPU, or paying per token.
Live status for the hosted APIs you might use instead of running this locally.
