ClapGenVideo motion transfer starts with a source performance.
Published by ClapGen · Updated
A source clip supplies visible movement and scene timing that a still image cannot provide. ClapGen Hypnosis combines that clip with replacement reference images and optional written direction. The public release currently uses simulated requests. Live generation and paid subscriptions are unavailable.
Decide whether you need motion transfer
Use this approach when an existing performance or scene is the starting point: a gesture, a walk, an interaction or the movement of an object. The source video and the replacement appearance serve different roles.
If you want to invent a scene from words, begin with text-to-video. If you only want a still composition to move, begin with image-to-video. A source performance is useful when the recorded action matters.
Prepare the current Hypnosis inputs
The current Hypnosis workflow requires exactly one source video between 4 and 30 seconds long, with readable width and height, plus 1 to 29 reference images. The source video and images together cannot exceed 30 media items.
Choose a clip where the relevant subject and action are visible. Large occlusions, rapid cuts or a subject leaving the frame make it harder to judge what should be preserved. Trim and review the clip before using it if the original is outside the required duration.
Use references that clearly show the intended replacement appearance. They do not need to imitate the source background unless that background is part of your direction. Only supply material you are entitled to process.
Map the source subject to the replacement
For a source with several people or objects, name which one changes. Use stable visible descriptions such as the person on the left wearing a light jacket, rather than an ambiguous phrase such as replace them.
Example brief: In the source clip, replace only the person on the left with the adult fictional character in the first reference image. Keep the other person and the visible handoff of the notebook. Use the reference character's green jacket and dark trousers.
This is an illustrative transformation brief, not a tested result. A body part, garment or accessory from the source may no longer exist on the replacement; inspect those interactions instead of assuming the transformation can preserve every contact unchanged.
Check direction, framing and duration together
Hypnosis offers motion and object-replacement directions. Select the direction that matches what you intend to change, and explain important identities or preserved objects in the optional prompt when needed.
The workflow follows the inspected source duration and uses adaptive framing. It is not a general request for an arbitrary new clip length or a promise of a particular crop. Follow the interface messages for the media actually supplied.
An input being accepted in the current demo does not prove that a future live model can produce the requested transformation.
Inspect the result as a sequence of interactions
When reviewing a future live result, check the beginning, key actions and ending. Look at which hand touches an object, how feet contact the ground, whether a held item stays attached and whether the replacement remains in the correct position.
If several subjects are present, compare each mapping separately. Check that an unchanged person was not replaced, a replacement did not switch positions and a face or outfit did not change partway through the clip.
A motion-transfer request is not a guarantee of an exact frame-for-frame copy, perfect identity or physically correct movement. A clear mapping makes errors easier to spot and a revised brief easier to explain.
Distinguish Hypnosis from Motion Control
Video Studio also has a Motion Control workflow for compatible models. Its source-media requirements and controls are defined by that model. Do not apply Hypnosis's image count or duration limits to every motion model.
For example, a workflow may ask for a subject image and a driving video rather than the Hypnosis replacement-reference arrangement. Choose the workflow first and read its input messages before assembling the files.