3 ms·
Or even just breaking them down into smaller chunks (prioritise loading the ones closer to where the user is looking) could help
by promiseofbeans 3y ago
Or even just breaking them down into smaller chunks (prioritise loading the ones closer to where the user is looking) could help
- duckworthd 3y agoThe viewer biases towards assets closer to user's camera (otherwise you'd have to load the whole scene!). We tried training SMERF with a larger number of smaller submodels, but at some point, it becomes too onerous to train and quality begins to suffer.