EasyAnimateV5 Sampler for Video to Video

EasyAnimateV5_V2VSampler is the most capable node in the EasyAnimate pack, and the one that justifies loading a -Control or -Control-Camera checkpoint at all. It's video-to-video, sure, but the interesting part is the optional inputs: control video, a reference image, and camera conditions. Wire all of them up and you're doing controlled, camera-directed video generation - the V5.1 headline features - in a single sampler.

It inherits the V2V flow from the legacy sampler but adds the V5 layer: Flow scheduler as the default, TeaCache on, and four optional inputs that unlock the control pipeline. This node only behaves like the demos if your loaded model matches what you're asking it to do, so the model wiring matters more here than anywhere else in the pack.

The inputs that matter

• easyanimate_model - from the loader. This is the node where model_type really counts. Inpaint weights use validation_video for restyling; Control weights use control_video (plus the extras below). Load EasyAnimateV5.1-12b-zh-Control or the -Control-Camera variant and set model_type to Control.
• validation_video (optional) - your source clip, for Inpaint models.
• control_video (optional) - a conditioning video (Canny, pose, depth) for Control models. This is also where a trajectory video from CreateTrajectoryBasedOnKJNodes goes.
• camera_conditions (optional) - a STRING, force-input. This is the wire that the camera stack feeds: CameraTrajectoryFromChaoJie's camera_trajectory output plugs straight in here. Requires the -Control-Camera weights.
• ref_image (optional) - a reference first frame for Control models (trajectory and camera workflows use this to anchor the shot).
• denoise_strength - 0.05–1.0, default 0.7. How much source structure survives.
• video_length - 1–49, step 4, default 49.
• scheduler - defaults to Flow. teacache_threshold 0.1 and enable_teacache true round out the V5 trio.

The output

images - the generated IMAGE frame batch, straight to VideoHelperSuite for saving.

How the workflows stack

CameraTrajectoryFromChaoJie (camera_trajectory) ─┐
                                                   ├─► EasyAnimateV5_V2VSampler
CreateTrajectoryBasedOnKJNodes (image)            ─┘   (camera_conditions + control_video + ref_image)

The official V5.1 control-camera and control-trajectory workflow JSONs live in the pack's comfyui/v5.1/ folder if you want the whole graph without wiring it by hand.

Notes

This node is simultaneously the pack's high point and its most brittle corner. When everything matches - -Control-Camera weights, Control model type, trajectory length equal to video_length - the camera control is genuinely fun. When it doesn't, you get drift, ignored conditions, or the model refusing to move at all. The community enthusiasm around these features peaked at V5.1's launch in early 2025, and the pack's momentum faded as Wan and Hunyuan took over - so treat this as a capable, somewhat experimental control surface rather than a polished flagship. The 12B is also VRAM-hungry, so keep the qfloat8 offload mode and the 7B in mind for your first passes.