From 227444a6b4f16b29e1c8337563a1bae99b2ad943 Mon Sep 17 00:00:00 2001
From: aiihn <101506918+aiihn@users.noreply.github.com>
Date: Wed, 25 Mar 2026 06:28:07 +0800
Subject: [PATCH] Enhance HunyuanWorldPlayOperator and Pipeline with
 configurable motion parameters

- Added forward_speed, yaw_speed_deg, and pitch_speed_deg as parameters to HunyuanWorldPlayOperator and HunyuanWorldPlayPipeline.
- Updated the initialization of HunyuanWorldPlayOperator to accept new parameters.
- Modified the pipeline to pass motion parameters to the operator and allow runtime adjustments.
- Updated tests to include new parameters for consistency in behavior.
---
 .../operators/hunyuan_worldplay_operator.py   | 19 +++++++++++++++----
 .../pipeline_hunyuan_worldplay.py             | 19 ++++++++++++++++++-
 test/test_hunyuan_worldplay.py                |  3 +++
 3 files changed, 36 insertions(+), 5 deletions(-)

diff --git a/src/openworldlib/operators/hunyuan_worldplay_operator.py b/src/openworldlib/operators/hunyuan_worldplay_operator.py
index c8c97c03..562b38dc 100644
--- a/src/openworldlib/operators/hunyuan_worldplay_operator.py
+++ b/src/openworldlib/operators/hunyuan_worldplay_operator.py
@@ -15,10 +15,21 @@
 
 
 class HunyuanWorldPlayOperator(BaseOperator):
-    def __init__(self, operation_types=None, interaction_template=None):
+    def __init__(
+        self,
+        operation_types=None,
+        interaction_template=None,
+        *,
+        forward_speed: float = 0.08,
+        yaw_speed_deg: float = 3.0,
+        pitch_speed_deg: float = 3.0,
+    ):
         if operation_types is None:
             operation_types = ["action_instruction"]
         super().__init__(operation_types=operation_types)
+        self.forward_speed = forward_speed
+        self.yaw_speed_deg = yaw_speed_deg
+        self.pitch_speed_deg = pitch_speed_deg
         self.interaction_template = interaction_template or [
             "forward",
             "backward",
@@ -117,9 +128,9 @@ def _is_action_sequence(self, interaction) -> bool:
         return all(item in self.interaction_template for item in interaction)
 
     def _actions_to_pose_json(self, actions: list[str]) -> dict:
-        forward_speed = 0.08
-        yaw_speed = np.deg2rad(3)
-        pitch_speed = np.deg2rad(3)
+        forward_speed = self.forward_speed
+        yaw_speed = np.deg2rad(self.yaw_speed_deg)
+        pitch_speed = np.deg2rad(self.pitch_speed_deg)
         motions: list[dict] = []
         for action in actions:
             move = {}
diff --git a/src/openworldlib/pipelines/hunyuan_world/pipeline_hunyuan_worldplay.py b/src/openworldlib/pipelines/hunyuan_world/pipeline_hunyuan_worldplay.py
index 02d98757..980a938b 100644
--- a/src/openworldlib/pipelines/hunyuan_world/pipeline_hunyuan_worldplay.py
+++ b/src/openworldlib/pipelines/hunyuan_world/pipeline_hunyuan_worldplay.py
@@ -38,6 +38,9 @@ def from_pretrained(
         overlap_group_offloading: bool = True,
         init_infer_state: bool = True,
         infer_state_kwargs: Optional[dict] = None,
+        forward_speed: float = 0.08,
+        yaw_speed_deg: float = 3.0,
+        pitch_speed_deg: float = 3.0,
         **kwargs
     ) -> 'HunyuanWorldPlayPipeline':
         """
@@ -90,7 +93,11 @@ def from_pretrained(
             action_ckpt=model_path,
             **kwargs
         )
-        operators = HunyuanWorldPlayOperator()
+        operators = HunyuanWorldPlayOperator(
+            forward_speed=forward_speed,
+            yaw_speed_deg=yaw_speed_deg,
+            pitch_speed_deg=pitch_speed_deg,
+        )
         
         return cls(
             synthesis_model=synthesis_model,
@@ -165,6 +172,9 @@ def __call__(
         model_type: str = "ar",
         user_height: Optional[int] = None,
         user_width: Optional[int] = None,
+        forward_speed: Optional[float] = None,
+        yaw_speed_deg: Optional[float] = None,
+        pitch_speed_deg: Optional[float] = None,
         **kwargs
     ):
         """
@@ -194,6 +204,13 @@ def __call__(
         Returns:
             HunyuanVideoPipelineOutput: 包含生成的视频帧
         """
+        if forward_speed is not None:
+            self.operators.forward_speed = forward_speed
+        if yaw_speed_deg is not None:
+            self.operators.yaw_speed_deg = yaw_speed_deg
+        if pitch_speed_deg is not None:
+            self.operators.pitch_speed_deg = pitch_speed_deg
+
         video_length = num_frames
         pose_value = interactions if interactions is not None else pose
         if pose_value is None:
diff --git a/test/test_hunyuan_worldplay.py b/test/test_hunyuan_worldplay.py
index 4de39184..616f9dec 100644
--- a/test/test_hunyuan_worldplay.py
+++ b/test/test_hunyuan_worldplay.py
@@ -21,6 +21,9 @@
     prompt=prompt,
     image_path=image_path,
     interactions=interaction_signal,
+    forward_speed=0.08,
+    yaw_speed_deg=3.0,
+    pitch_speed_deg=3.0,
 )
 
 save_video_path = os.path.join(output_path, "hunyuan_worldplay_demo.mp4")