- Always. Its Best Practice.
- what? Go google what a light map is, see if the results help?
- you probably can’t merge a whole castle into a single thing. Look up what Texel Density is, understand it. Then partion off the model in a way that makes the most sense in order to preserve that; Resolution wise, each piece at around 1064 is already going to somewhat cost you a few hours to bake, but the shadows are bound to look crispier than they would at say 256. Except this depends on the texel density and the size of the pieces as well.
You probably need to spend another month building the castel if you werent aware of how lightmap UVs work. And that’s perfectly OK. Do it right, instead of taking shortcuts.
And on that note, you want to pixel align to the center of the pixel for all lightmap UVs.
And you want the pieces to always fall at a consistent pixel defined distance so that your MIP wont affect the quality too much when you apply different scale values to the lightmap - all always in power of 2 format.
There are some obvious issues with small stuff.
Overall:
If you build your UV on 256, correctly, it will scale up nicely to anything ^2 above it.
If you build your UV on 526 or higher ^2 value, there is no absolute guarantee it will look nice as you scale down.
So, for models you want to use in a videogame, see if you can get away with 256 and extremely simple lightmap UVs.
Having done this on several castles, it’s not impossible if you consider small 2m by 2m sections.
Be mindful of slants, as they will go jaggy regadless of what you do. A proper lightmap UV shouldn’t really have any slants when possible; so a beam that unwraps at a 45deg angle is cut and rotated to be at 90deg instead - but an arch is an arch and you have to accept the jaggies will exist on it.