Windflow

Echo-Memory

JD Joy Future Academy · Released 6 June 2026

An action-conditioned world model study from JD that compares four kinds of memory for keeping a generated scene consistent when the camera leaves and comes back. It is built on Wan 2.1 1.3B and released with code, checkpoints and ComfyUI nodes under CC BY 4.0.

Try on Hugging Face SpaceGet the weightsCan I use it commercially?
Weights
Open weightsCC BY 4.0
Commercial use
Permissive license
Real-time
No real-time claim
Where to try
1 placeHugging Face Space

01 Controls

What you control

  • A first frame to start from
  • A text prompt
  • Camera actions for each chunk of video

02 License

Can you use it commercially?

Permissive license

The weights' model card declares CC BY 4.0, a permissive license that allows commercial use. Attribution is required under CC BY 4.0. The checkpoints are fine-tunes of Wan 2.1 1.3B, which is Apache-2.0.

Declared license: CC BY 4.0 · Model card

A reading of the published license text, not legal advice.

03 Speed

Does it run in real time?

No real-time claim. The cited sources make no real-time claim.

Figures are the authors' own, with their settings and hardware. They are not comparable across models.

04 Setup

How to run it

  1. 01Download the Wan2.1-T2V-1.3B base model and one of the Echo-Memory checkpoints (context, compression or state-space memory) from Hugging Face.
  2. 02Point the inference script at the base model and a first frame, then generate the video chunk by chunk with camera actions.
  3. 03ComfyUI users can install the project's custom nodes (loader, camera action, generate) from its comfyui folder or the Comfy Registry.

Summarised from the project's own instructions. Follow the source for exact commands and versions.

05 Access

Where to try it

Listed alphabetically. Windflow maintains this directory and is one of the hosts; a listing does not promise capacity at any given moment.

06 Questions

Echo-Memory FAQ

Is Echo-Memory open source?

Echo-Memory publishes downloadable weights under CC BY 4.0. The declared license is permissive.

Can I use Echo-Memory commercially?

The weights' model card declares CC BY 4.0, a permissive license that allows commercial use. Attribution is required under CC BY 4.0. The checkpoints are fine-tunes of Wan 2.1 1.3B, which is Apache-2.0.

Does Echo-Memory run in real time?

The cited sources do not claim real-time operation.

Where can I try Echo-Memory?

You can try it on Hugging Face Space.

What does Echo-Memory take as input?

You give it a first frame to start from, a text prompt and camera actions for each chunk of video.

How do I run Echo-Memory?

Following the official instructions: Download the Wan2.1-T2V-1.3B base model and one of the Echo-Memory checkpoints (context, compression or state-space memory) from Hugging Face. Point the inference script at the base model and a first frame, then generate the video chunk by chunk with camera actions. ComfyUI users can install the project's custom nodes (loader, camera action, generate) from its comfyui folder or the Comfy Registry.

What are alternatives to Echo-Memory?

Similar models in this directory include ABot-World (open weights, permissive license), AlayaWorld (open weights, custom terms) and DreamX-World (open weights, permissive license).

Similar world models

All 56 in the directory