AI Video Generation with Start & End Images

Upload a first frame and a last frame; the AI generates the motion in between. Put spoken lines in double quotes.

First frame

First frame No file selected

Last frame

Last frame No file selected

Motion description

Auto-direct a continuous sequence of several distinct actions — no fewer than three, no more than five — that begins exactly at the pose in Picture 1 and finishes exactly at the pose in Picture 2.Compare the two pictures before deciding anything. It is the same subject wearing and carrying the same things, so say that first and say it once. Then go through the differences one at a time: which foot the weight is on, where each limb has got to, how the head and hips have turned, what the hands or paws are holding and where they are pointing, what the face is doing, and how much the loose things in the scene — dust, leaves, water, hair, cloth — have been stirred.Then plan the chain of actions that gets from the first pose to the second. Each action must be a complete, recognisable move on its own — a step, a turn, a reach, a spin, a lunge, a hop, a crouch, a gesture, a pivot — not a sub-phase of one bigger motion. Consecutive actions must differ from each other in at least one of: direction of travel, which limb leads, tempo, or level (high/low). The intermediate actions pass through poses that appear in neither picture; only the final action is allowed to arrive at Picture 2. Build every action out of body mechanics: what pushes off, what the weight travels through, which joint opens first, which limb leads and which arrives late, where the momentum overshoots before it settles. Choose a different chain each time.Carry each action in three stages: the load, where the body gathers and the ground takes the weight; the drive, where the movement is fastest and the frame is hardest to read; and the arrival, where it runs out. The arrival of one action is the load of the next — no freeze, no dead frames between them; the body is always either settling or gathering. Vary the tempo across the chain: at least one action is quick and sharp, at least one is slow and weighted. Spread the actions across the full duration so the last arrival lands on Picture 2 exactly at the end, not early. Nothing turns into anything else at any point — every difference between the pictures is reached by the same things moving.Let the scene answer to every action, not just the last: whatever each move throws, drags, displaces or disturbs does what it would really do, and the shadows and reflections travel with the body through every action. Give each action its own sound, distinct from the others. Background music IS requested for this video: describe it in non_diegetic_music (instruments, tempo, rhythm, dynamics), and let its beat line up with the drive of each action.

2584/5000

Video length:

10s

15s

20s

25s

Aspect ratio:

9:16 (Portrait)

TikTok / Reels

16:9 (Landscape)

YouTube

Video Demo