"Pose matching" isn't a feature Pose AI ships, and it isn't standard across mainstream AI photo generators either — there's no widely available capability that takes an uploaded reference photo and copies its exact pose onto a different person's still image. What is real, and worth understanding separately, is style-based pose selection for stills and Motion Control for video, which lets you reference existing motion in a way stills currently can't.
- For still images: Pose AI doesn't copy an exact pose from an uploaded reference photo. You pick a style or scene, or describe the pose in text, and identity-locked Nano Banana 2 generates your face into it.
- For video: Pose's Motion Control lets you direct camera movement (via Kling) and subject motion (via SeedDance), and SeedDance's multimodal input can reference up to three existing video clips — real reference-based capability that stills don't currently have.
- The difference between the two is worth being precise about: prompt/style control for photos is not the same thing as reference-based motion for video, and neither is branded "pose matching."
Motion Control (real)
Motion Control is a real feature in Pose's Video Studio. Kling handles camera movement — pans, pushes, orbits — and SeedDance handles subject motion, including motion transfer, where the movement in an existing reference clip is applied to your generated subject. SeedDance's multimodal reference input accepts up to three video clips as reference material, which means video generation genuinely can draw on existing motion in a way that goes beyond a text prompt.
Reference-photo pose copying for stills (not currently supported)
For still photos, there's no equivalent. Pose does not support uploading a reference photo and having the exact pose — the arm position, the angle, the stance — copied onto your identity-locked face. Pose selects on scene/style, or a text description in prompt mode, not on an uploaded second image's pose. If this is a hard requirement for your workflow, it's worth knowing that limitation up front rather than assuming it works like video reference does.
Which tools support pose reference, and how
| Medium | Reference support | How it actually works in Pose |
|---|---|---|
| Still photos | No exact-pose reference | Style/scene selection or text description; identity read from one selfie |
| Video | Yes, for motion | SeedDance references up to 3 clips; Kling controls camera movement via Motion Control |
Video generation in Pose has genuine reference-based capability that still-image generation doesn't currently have — the two shouldn't be described as the same feature.
See Motion Control in Pose's native video generation.
Browse Pose's photo styles for how pose is set on stills today.