Skip to content

Decentraland DAO inches closer to integrating AI-generated scenes

Decentraland DAO has approved a grant proposal to fund the development of DCLBuilderAI, an artificial intelligence tool that creates scenes and other assets through natural language prompts.

A grant proposal authored by Adi sought $56,000 in funding from the Decentraland DAO with an initial six-month roadmap. The DAO approved the proposal on November 29th, 2022, with 97 votes in favor (3,375,150 voting power) and 8 votes against (533,914 voting power).

The grant will fund the development of DCLBuilderAI. This artificial intelligence tool will be able to create scenes and other assets through natural language prompts. DCLBuilderAI will provide a multi layout-view displaying both the code for the scene and the preview.

The funding breakdown is as follows:

OpenAI Project Community Access to DCL- $8,000
Web Tooling, Infra for frontend and backend, maintenance of project- $12,000
Salaries- $32,000
Misc expenses- $4,000

The project will be open-source, and a core part of the process will be allowing the Decentraland community to experiment with the tool. However, when the $8,000 in funding runs dry, end users will have to pay for the service.

The team behind the project consists of two full-stack software engineers, Aditya and Tave, who have experience with AI-based projects and have already successfully delivered projects within Decentraland.

The community reacted positively, with DAO members broadly in assent. DAO member allvintage commented on the potential for this project to push Decentraland forwards:

This is a fantastic proposal. It could help get more scenes up & running on lots of currently empty LANDs. This extends DCL capabilities and even has the potential to be fully integrated. I would be a heavy user of this.

DCLBuilderAI has already released two monthly updates, and the building process is well underway.

In the first update, the team announced they had created a data set with 100 examples and had already seen improvements in code generation. However, they reported an issue speculating that perhaps the base model had been too heavily trained on outdated Decentraland. They announced their plan to increase the data set to 500 examples to prevent failed queries.

In the second update, the grant changed its roadmap targets, outlining a nine-month roadmap instead of six at the advice of the DAO Grant support squad. They have a basic front end ready and are working on launching the model on the 22nd of February and have code generation working simultaneously. They stated the next steps would be to have a shareable public data set, make code generations work in February, and hire a data scientist.

The project is on-track, and ostensibly in the next few months, Decentraland users will be able to use DCLBuilderAi to create imagery and scenes with language prompts.