Introducing ``Fill 3D,'' which uses image generation AI to load a photo of a room and add your favorite furniture without any failure just by inputting a prompt

' Fill 3D ' was announced, which uses image generation AI to generate furniture within a room photo. Fill 3D understands the three-dimensional structure and lighting of the room shown in the photo, and can generate furniture exactly as input at the prompt anywhere on the photo.

Fill 3D

You can understand what kind of AI service Fill 3D is by watching the movie below.

'Fill 3D', which uses image generation AI to draw 3D furniture according to the structure and lighting of the room in the photo, looks like this - YouTube

Load the room photo into Fill 3D.

Drag to select a corner of the room.

Enter 'pink bed' in 'Prompt' in the right column.

Furthermore, select the area above the selection separately.

Now enter the prompt 'red painting'. Turn on the 'Sticky' switch to generate objects along the wall and click 'Stage'.

Then, a bed and a painting were generated on the photo as shown below. You can see that the bed and painting are not just drawn casually, but are properly generated to match the perspective of the photo so that there is no disruption. In addition to each item, shadows are also drawn to match the light coming in from the window.

The image generated by Fill 3D is the same as the loaded photo, so the resolution will not be 512 x 512 pixels. It takes less than 1 minute to generate, and even high-resolution images can be generated in 1 to 2 minutes.

Fill 3D is an AI company > Provided by Function AI for a fee. One credit is required for each image to be loaded, and 5 credits costs $20 (about 3,000 yen), 20 credits costs $60 (about 9,000 yen), and 50 credits costs $100 (about 15,000 yen).

In addition, a simplified version of Fill 3D has been published on GitHub below, and can be used by entering the API key provided by Function AI.

GitHub - fill3d/fill: Generative fill in 3D.

in Software,   Web Service,   Design, Posted by log1i_yk