From 632bef29a617603e7a46441bafc08086ccfa09d2 Mon Sep 17 00:00:00 2001
From: Tobias Juelg <tobias.juelg@utn.de>
Date: Sun, 10 May 2026 10:30:58 +0200
Subject: [PATCH] feat(convert): lerobot data convert video codec

---
 python/rcs/__main__.py                | 4 ++++
 python/rcs/lerobot_joint_converter.py | 4 ++++
 2 files changed, 8 insertions(+)

diff --git a/python/rcs/__main__.py b/python/rcs/__main__.py
index dc0b3cdf..344060d4 100644
--- a/python/rcs/__main__.py
+++ b/python/rcs/__main__.py
@@ -147,6 +147,9 @@ def lerobot_convert(
     success: Annotated[bool, typer.Option(help="Only include successful episodes. Example: --success")] = True,
     n: Annotated[int, typer.Option(help="Maximum number of episodes to convert. -1 means all. Example: --n 50")] = -1,
     video_encoding: Annotated[bool, typer.Option(help="Should the image data be video encoded")] = False,
+    video_backend: Annotated[
+        str | None, typer.Option(help="Video backend to use if image data is video encoded e.g. torchcodec")
+    ] = None,
 ):
     cameras = camera_specs_to_configs(camera_specs) if camera_specs is not None else list(DEFAULT_CAMERAS)
     run_conversion(
@@ -164,6 +167,7 @@ def lerobot_convert(
         success=success,
         n=n,
         video_encoding=video_encoding,
+        video_backend=video_backend,
     )
 
 
diff --git a/python/rcs/lerobot_joint_converter.py b/python/rcs/lerobot_joint_converter.py
index 33d0fced..60b800c3 100644
--- a/python/rcs/lerobot_joint_converter.py
+++ b/python/rcs/lerobot_joint_converter.py
@@ -99,6 +99,7 @@ def __init__(
         image_batch_size: int = DEFAULT_IMAGE_BATCH_SIZE,
         per_robot_arm_dim: int = DEFAULT_PER_ROBOT_ARM_DIM,
         video_encoding: bool = False,
+        video_backend: str | None = None,
     ):
         self.root = Path(root)
         self.conn = duckdb.connect()
@@ -133,6 +134,7 @@ def __init__(
             features=self._build_features(),
             image_writer_threads=10,
             image_writer_processes=5,
+            video_backend=video_backend,
         )
 
     def _build_features(self) -> dict[str, dict[str, Any]]:
@@ -416,6 +418,7 @@ def run_conversion(
     success: bool = True,
     n: int = -1,
     video_encoding: bool = False,
+    video_backend: str | None = None,
 ) -> None:
     robot_type_converted = RobotType(robot_type)
     gripper_type_converted = GripperType(gripper_type)
@@ -432,6 +435,7 @@ def run_conversion(
         image_batch_size=image_batch_size,
         per_robot_arm_dim=per_robot_arm_dim,
         video_encoding=video_encoding,
+        video_backend=video_backend,
     )
     converter.generate_examples(success=success, n=n)