Pose Map (Video)
Turns a video into a pose skeleton video (DWPose): the joints of every person in the clip, drawn on black. Feed the result to Seedance 2.5 as the reference video to transfer a person's motion without their body shape. Better than a depth map when the new character's build differs from the source, or for big gestures and dances; a depth map carries the source body's outline. Clips up to 15 seconds (and at most 900 frames) as a regular MP4 or MOV. Inside a blueprint, keep clips to about 5 seconds at 30 fps: a blueprint stops waiting after 10 minutes, and fal sometimes runs this model slowly. Charged up front at the 60 fps rate; the hold is normally refunded down to the frames in the returned video.
Pose Map (Video)
Node type: video:dwpose_video
Category: video
Description
Turns a video into a pose skeleton video (DWPose): the joints of every person in the clip, drawn on black. Feed the result to Seedance 2.5 as the reference video to transfer a person's motion without their body shape. Better than a depth map when the new character's build differs from the source, or for big gestures and dances; a depth map carries the source body's outline. Clips up to 15 seconds (and at most 900 frames) as a regular MP4 or MOV. Inside a blueprint, keep clips to about 5 seconds at 30 fps: a blueprint stops waiting after 10 minutes, and fal sometimes runs this model slowly. Charged up front at the 60 fps rate; the hold is normally refunded down to the frames in the returned video.
Pricing
- Cost: ~16 credits per seconds
- Per-second rate: The figure is the rate at a 5-second reference run. Our margin is progressive and steepest on the smallest runs, so a clip shorter than that can cost more per second, and a model priced with a flat multiplier is exact at every length. The charge is always computed from the real duration.
- Notes: fal bills this model by compute time ($0.0006 per compute second), and the time per frame varies a lot between runs: on 2026-10-03 the same 90-frame clip took between 7 and 219 compute seconds. Priced at $0.0015 per frame, about the slowest rate seen. Clips are limited to 15 seconds and 900 frames, counted from the file before the run; a longer clip, a fragmented MP4, or one whose frames cannot be counted is refused before you are charged. The run is charged up front for the counted frames at the 60 fps rate, then normally refunded down to the frames in the pose video fal returns, so a 30 fps clip pays for its own frames.
Canvas ports
These appear as port handles on the left side of the node.
| ID | Label | Details |
|---|---|---|
video_url |
Video | VIDEO (required) — Source clip with the person whose motion you want, up to 15 seconds (and at most 900 frames) as a regular MP4 or MOV. Trim it first if it feeds Seedance 2.5, which takes its length from the reference. |
Sidebar config
These render as form fields in the right-side config panel when the node is selected.
| ID | Label | Details |
|---|---|---|
draw_mode |
Draw Mode | TEXT · options: full-pose, body-pose, face-pose, hand-pose, face-hand-mask, +2 — What to draw. Body pose (the default) is the one to feed Seedance for motion. |
Outputs
| ID | Label | Type |
|---|---|---|
video |
Pose Video | VIDEO |
What it is for
Motion transfer for people. Wire in a clip of someone moving and you get a skeleton video back: colored lines for the joints of every person in the frame, on black. The face, the clothes, the room and the shape of the body are gone. Wire that output into Seedance 2.5's Reference Video port next to your character images, and say what it is in the prompt ("@Video1 is a pose skeleton, take only the motion and timing from it").
Pose or depth
Use Pose Map when the new character's build is different from the person in the source, or for big gestures and dances. A depth map carries the source body's outline (head size, build), so the take drifts toward that body. Use Depth Map when you also want the camera move and the shapes of the room, which a skeleton does not carry.
Body pose, the default Draw Mode, is the one to feed Seedance. Full pose adds the hands and the face.
Cost in dollars
fal bills this model by compute time, and the time per frame swings a lot from run to run, so the price is set at about the slowest rate measured: $0.0015 per frame. A 10 second clip at 30 fps (300 frames) costs $0.45. The run is charged up front at the 60 fps rate, then refunded down to the frames fal returns.
Length and speed
Clips can be up to 15 seconds and 900 frames; trim longer ones first. Speed varies a lot: the same 3 second clip took about 4 minutes on one run and about 10 seconds on another. The 15 second limit is what lets even the slowest runs finish before the run stops waiting.
Inside a blueprint, keep clips to about 5 seconds at 30 fps for now. A blueprint stops waiting after 10 minutes, which a slow run of a longer clip can pass, and then the steps after this one do not run. This holds until blueprints give slow nodes a longer clock.
Auto-generated from the Wireflow node registry.