How to Make a Great Start Frame for Kling and Veo
The most reliable way to get good video out of image-to-video tools like Kling and Veo is not prompting — it is the quality of your start frame. Video models struggle to create what isn't in the first frame: start from a blank expression and you get a blank-faced video. Give the model a frame that already contains the emotion, cinematic composition, and lighting direction, and it only has to add motion — results improve dramatically.
Why the start frame is the ceiling
Image-to-video models are trained to preserve the first frame. Face, framing, lighting, and color largely carry through; what the model adds is movement over time.
That's why asking the motion prompt to create an emotion that isn't in the frame ('make her smile') tends to produce mushy, uncanny transitions. Emotion belongs in the frame; motion belongs in the prompt.
Four ingredients of a strong start frame
① Emotion already on the face — eyes brimming with tears, a face about to break, a laugh about to burst. Escaping the AI default of 'neutral face, frontal view' is half the job.
② Cinematic composition — a deliberate shot size (close-up, medium), a purposeful camera angle, negative space.
③ Room to move — the moment just before the peak of an action, not the peak itself. Mid-turn, mid-reach: it tells the model where the motion goes.
④ Directional lighting — a clear key-light direction helps the model keep lighting logic consistent through the motion.
The workflow: acting shots → start frames
Step 1 — Generate 15 acting cuts from one actor photo: Cinema Forest's Acting Shots produces 15 performance frames per emotional series (inner turmoil, breakdown, solitude, power, allure, romance) with the same face throughout — a supply of frames with the emotion already in them.
Step 2 — Pick the cut just before the emotional peak, where motion has somewhere to go.
Step 3 — Reframe with Multi Angles if needed: re-shoot the same person and emotion from 15 camera angles.
Step 4 — Feed the video tool: start frame + a motion prompt that only directs camera and action. Cinema Forest's prompt tools format motion prompts for Kling, Seedance, and Veo/Flow.
FAQ
What resolution should a start frame be?
2K at 16:9 (2752×1536) is plenty. Most video tools accept 1080p–2K input, and composition and emotion matter far more than extra pixels.
Can't I just prompt the expression?
Creating an emotion that isn't in the start frame tends to produce awkward transitions. Put the emotion in the frame; keep the prompt focused on camera and action.
Do I need an end frame too?
Where supported (e.g., Kling), a start + end frame pair gives the motion a destination and stabilizes it. Using another cut of the same character as the end frame requires face consistency across cuts.