Translation depth controls whether Node to Code follows calls to other functions or macros and includes their graphs in the translation. The default is 0, which translates only the graph you triggered the translation from.
How depth works
When your graph calls a function or macro defined in the same Blueprint, Node to Code can optionally follow that call and translate the inner graph as well. Each level of nesting adds one to the depth count.
| Depth | What gets translated |
|---|
0 | Only the selected or active graph |
1 | The active graph plus any functions/macros it calls directly |
2 | The above, plus any functions/macros those call |
3–5 | Each additional level follows one more layer of calls |
Configuring translation depth
Open Project Settings
Go to Edit → Project Settings in the Unreal Engine menu bar.
Navigate to Node to Code
Select Plugins → Node to Code from the left sidebar.
Set the depth
Under Code Generation, change Max Translation Depth to a value between 0 and 5.
Example
Suppose you have a function HandleDamage that calls two helper functions: ApplyStatusEffect and UpdateHealthUI.
- At depth 0, only
HandleDamage is translated. The calls to the helpers appear as function calls in the output, but their implementations are not included.
- At depth 2,
HandleDamage, ApplyStatusEffect, and UpdateHealthUI are all translated. If ApplyStatusEffect itself calls another function, that would also be included.
Higher translation depth significantly increases the number of tokens sent to the LLM. Each additional level can multiply the input size depending on how many nested calls exist. Watch your costs carefully when increasing depth.
TranslationDepth puts it plainly: “This setting can greatly impact costs and context window utilization, so be mindful!”
Some LLMs also have context window limits. Very deep translations of complex Blueprint hierarchies may exceed those limits and fail or produce incomplete output.
Recommended approach
Start at depth 0. Only increase depth when you specifically need to understand or reimplement the full call hierarchy. Depth 1 is often enough to get the context you need without a large cost increase.
