AI video concepts and motion references
Use a concept workflow when the story or visual effect matters more than identical movement. Use a motion reference when separate dancers, turns and choreography need a shared sequence. Our examples show both approaches, with different limits.
A concept describes the events
The marigold path scene specifies a walk, a crouch and a handful of petals. It does not need a video of the same person doing those actions first. The ghost-bride transformation and pet bridge also use scene direction and identity references. A prepared scene image can establish the setting when the story needs particular props.
A concept leaves timing to the model
My Ride Is Here preserves the broom takeoff and flight away, but the generated camera uses a windshield viewpoint. Ramalama keeps one dog and three costumes, yet the changes do not land on the exact beats of a reference clip. These differences matter if you want an identical shot or music-synchronized edit.
A motion reference supplies a shared performance
The two-person Thriller and three-cat La Calabaza samples use generated motion studies before identity replacement. Each dancer has a stable role, even during turns. The final result should preserve different movements for different dancers; assigning two identities is not enough if both are made to copy one track.
Compare the reference and final result
Play the generated motion study and final dance from the start. Check dancer count, individual arm and leg movement, accessories, camera framing and the ending. La Calabaza’s right-hand cat is partly out of frame near the end. The two-adult Thriller sample has no zombie makeup; it does not establish a pet version.
Choose by what must stay fixed
If your priority is a pet appearing on a glowing bridge, the concept’s setting and approach may be sufficient. If your priority is a recognizable dance with two independent performers, the movement reference carries more of the requirement. Results can vary with your photos. The template form shows its supported format and photo slots; watch the whole result before sharing.
Questions before you create?
Does every AI video template need a video master?
No. The nine concept examples were made without a separate motion master. The two dance examples use generated motion studies because their movement sequences matter.
Does motion guidance guarantee frame-perfect choreography?
No. Inspect the actual sample and full playback. A successful generation or matching duration alone does not establish movement accuracy.