diff --git a/CHANGELOG.md b/CHANGELOG.md
index 5c0e83d..910cfdb 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -9,6 +9,7 @@ and this project adheres to [Semantic Versioning](http://semver.org/spec/v2.0.0.
 
 - Change default temporal property to datetime
 - Enable Converter.columns list and tuple types
+- Add BaseConverter get_columns hook to customize columns after reading the file
 - Update STAC processing extension
 
 ## [v0.2.11] - 2025-10-09
diff --git a/vecorel_cli/conversion/base.py b/vecorel_cli/conversion/base.py
index 6070945..8185276 100644
--- a/vecorel_cli/conversion/base.py
+++ b/vecorel_cli/conversion/base.py
@@ -51,7 +51,7 @@ class BaseConverter(LoggerMixin):
     variants: dict[str, Sources] = {}
     variant: Optional[str] = None
 
-    columns: dict[str, str|Sequence[str]] = {}
+    columns: dict[str, str | Sequence[str]] = {}
     column_additions: dict[str, str] = {}
     column_filters: dict[str, Callable] = {}
     column_migrations: dict[str, Callable] = {}
@@ -83,6 +83,9 @@ def layer_filter(self, layer: str, uri: str) -> bool:
     def post_migrate(self, gdf: GeoDataFrame) -> GeoDataFrame:
         return gdf
 
+    def get_columns(self, gdf: GeoDataFrame) -> dict[str, str | Sequence[str]]:
+        return self.columns.copy()
+
     def get_cache(self, cache_folder=None, **kwargs) -> tuple[AbstractFileSystem, str]:
         if cache_folder is None:
             _kwargs = {}
@@ -299,7 +302,6 @@ def convert(
         original_geometries=False,
         **kwargs,
     ) -> str:
-        columns = self.columns.copy()
         self.variant = variant
         cid = self.id.strip()
         if self.bbox is not None and len(self.bbox) != 4:
@@ -340,6 +342,8 @@ def convert(
         gdf = self.migrate(gdf)
         assert isinstance(gdf, GeoDataFrame), "Migration function must return a GeoDataFrame"
 
+        columns = self.get_columns(gdf)
+
         # 2. Run filters to remove rows that shall not be in the final data
         gdf = self.filter_rows(gdf)