Changelog.ggBeta
HomeGamesUpdatesPricingMethodology
Steam News17 April 20262mo ago

Patch 2026.04.1: Image-to-Model

Hi everyone, This update took a bit longer than expected, but trust me this is a really fun one. The main focus of this update was image-to-model, and its something I've already spent hours playing around with.

Full notes

Full LAN Party update

Read the full published notes in a cleaner layout. The original post stays linked below.

What changed

5 fixes3 additions4 changes0 removals
  • UI and audio
  • Gameplay
  • Compatibility
  • Fixes
changedThis tool allows you to bring in any image and create a fully textured 3D model with a click of a button! We've been really pleased with our test results, and we've found that the output quality is much better than text-to-model. Any model that you create can be configured as any type of asset as well.
changedWe have noticed that as the output quality improves, the models take longer to generate. Previously, you would have to just sit and wait in the asset importer until the model was finished. To fix this, we implemented a persistent task queue that will execute the generation tasks in the background, meaning you can leave the importer and come back to import your generated models. It will also save all of your previous generations as well, so you can switch between models in the importer at any time. In addition, premium users can have up to 5 models generating at the same time, which I've used a lot to generate using the same image several times at once and then pick the best one at the end.
addedWe hope you enjoy this new feature as much as we have! We've got several other features and improvements that are nearing completion (controller support, revamped text chat, automatic avatar rigging), as well as several improvements and bug fixes, so keep your eyes peeled for future updates!
addedAdded image-to-model support.
addedAdded a persistent queue for model generation tasks.
changedImprovements:

LAN Party changes

changedThis tool allows you to bring in any image and create a fully textured 3D model with a click of a button! We've been really pleased with our test results, and we've found that the output quality is much better than text-to-model. Any model that you create can be configured as any type of asset as well.
changedWe have noticed that as the output quality improves, the models take longer to generate. Previously, you would have to just sit and wait in the asset importer until the model was finished. To fix this, we implemented a persistent task queue that will execute the generation tasks in the background, meaning you can leave the importer and come back to import your generated models. It will also save all of your previous generations as well, so you can switch between models in the importer at any time. In addition, premium users can have up to 5 models generating at the same time, which I've used a lot to generate using the same image several times at once and then pick the best one at the end.
addedWe hope you enjoy this new feature as much as we have! We've got several other features and improvements that are nearing completion (controller support, revamped text chat, automatic avatar rigging), as well as several improvements and bug fixes, so keep your eyes peeled for future updates!
addedAdded image-to-model support.
addedAdded a persistent queue for model generation tasks.

Hi everyone,

This update took a bit longer than expected, but trust me this is a really fun one. The main focus of this update was image-to-model, and its something I've already spent hours playing around with.

This tool allows you to bring in any image and create a fully textured 3D model with a click of a button! We've been really pleased with our test results, and we've found that the output quality is much better than text-to-model. Any model that you create can be configured as any type of asset as well.

Here are some examples of a few models I created today:

And here's what they look like in-room:

We have noticed that as the output quality improves, the models take longer to generate. Previously, you would have to just sit and wait in the asset importer until the model was finished. To fix this, we implemented a persistent task queue that will execute the generation tasks in the background, meaning you can leave the importer and come back to import your generated models. It will also save all of your previous generations as well, so you can switch between models in the importer at any time. In addition, premium users can have up to 5 models generating at the same time, which I've used a lot to generate using the same image several times at once and then pick the best one at the end.

We hope you enjoy this new feature as much as we have! We've got several other features and improvements that are nearing completion (controller support, revamped text chat, automatic avatar rigging), as well as several improvements and bug fixes, so keep your eyes peeled for future updates!

Features:

  • Added image-to-model support.

  • Added a persistent queue for model generation tasks.

Improvements:

  • Several UI functionality improvements.

Fixes:

  • Fixed issue where if the first person to start streaming in a room did not stream to a screen, other players in the room would not get their stream until they put their stream on a screen.

  • Fixed issue where generated models could have gargantuan file sizes.

  • Fixed avatar hitbox issue that could prevent climbing on ledges and inclines.

  • Fixed issue where certain assets could halt the progress of loading into a room.

  • Fixed issue where scrolling could both zoom the camera and scroll the UI if a menu is open.

Source

Steam News / 17 April 2026

Open original post

Changelog.gg summarizes and formats this update. How we read updates.