As discussed in #1026, we can ease the tuning by limiting the range as shown in #1026 (comment), at least we should provide an option.

The problem is how should the range be obtained. There're plenty of ways, the best approach seems to be including them in meta.json and provide an API from the engine side that GUI can access. However, that part seems to be done by the owner @Hiroshiba when he releases new models. Or is the building script already published somewhere? I'd like to know further about the building process at current time.

To prevent the trouble, generating them on the fly seems to be a choice as well. Therefore, there goes the other two ways:

1. Do it in the engine and provide a new API. The engine calculates range info on the fly.
2. Do it in the GUI utilizing the existing APIs. The engine calculates range info on the fly.

With the pre-compute method as the third choice,

3. Do it in the engine and provide a new API. The engine grabs the pre-computed range info from meta.json.

Please feel free to leave comment for these choices or give the instruction of modifying building process (if possible).