LESSON 3 · YOUR FIRST CLIP

Bringing the starting frame to life: step by step

In this lesson we'll build our first working clip from start to finish. We already have an image — the starting frame [Image 1] — and our task isn't to describe it again, but to define the motion that will grow out of it. We'll walk through six steps, and at the end assemble a full copyable prompt for a specific portrait.

Nothing runs without a starting frame
WAN 2.7 Spicy's image-to-video mode always animates from a single starting frame. No image — no clip: pure text-to-video doesn't work here; that requires a separate t2v tool. The very first thing we always do is load [Image 1].

Six steps to your first clip

  1. Load the starting frame [Image 1]. This is the first and only frame from which the animation will unfold. The model already sees it — so there's no need to re-describe "a girl with brown hair by the window."
  2. Choose duration and resolution. Duration is 5 / 10 / 15 seconds, resolution is 720p or 1080p (480p isn't supported on 2.7). For a first test, 10s / 1080p is convenient: long enough for the motion to "open up," but not so long that regenerating means a long wait.
  3. Decide what the subject DOES. You need a concrete action verb, not a static state. Not "a calm scene," but "slowly turns toward the camera and smiles." One clear action beats five blurry ones.
  4. Decide how the camera moves to match that action. The camera should support the subject's motion: a gentle push-in for a calm portrait, tracking for running, a pull-out or orbit for a large-scale scene. A static camera under a sharp action (and vice versa) looks bad.
  5. Add pacing and changes in light/scene. Set the rhythm (slow / steady / builds up / rapid) and what changes over time: soft light intensifies, the curtain sways, a glint passes across the glass.
  6. Generate, evaluate, fix with ONE lever. Look at the result and change one parameter at a time: first the action, then the camera, then the pacing. That way you can see exactly what made the difference.

What motion is made of

SUBJECT

Action as a verb

turns, smiles, rises, rolls forward. A concrete gesture, not "peaceful scene."

CAMERA

Movement to match the action

static / pan / tilt / push-in / pull-out / orbit / tracking — matched to the scene.

⏱
PACING + LIGHT

Rhythm and changes

slow / steady / builds up; light intensifies, a glint passes, an arc of "calm → peak → settling."

Breaking it down on a specific frame

Take [Image 1]: a portrait of a girl by a window, soft daylight, she's looking off to the side. We want a 10-second clip: the girl slowly turns toward the camera and gives a soft smile, the camera moves in a little (a gentle push-in), and the light grows warmer toward the finish. First — how not to do it.

A peaceful scene. A beautiful young woman with brown hair sitting by a window. Soft daylight. She is calm and pretty. Cinematic portrait, high quality, detailed.

This describes the frame, not the motion: the model already sees all of that. No verb, no camera, no pacing — the output is a nearly static image with random jitter.

[Image 1] The woman slowly turns her head toward the camera and gives a soft, natural smile. Her hair shifts gently with the movement. The camera performs a slow, subtle push-in toward her face. Pacing is calm and steady throughout the shot. The daylight through the window warms slightly as the shot progresses, adding a gentle glow to her skin. Motion starts quietly, settles as she meets the camera's gaze. 10 seconds, 1080p.

Action (turns + smiles), camera (slow push-in), pacing (calm, steady), light change (warms slightly), and a motion arc (starts quietly → settles). Everything is about motion — nothing extra about the frame we already see.

negative_prompt — as a separate field

The negative lives in a separate field, not inside the main prompt. For a calm portrait it's important to rule out facial distortions and morphing.

blurry, warped face, distorted features, extra limbs, morphing, flicker, jitter, unnatural motion, fast motion, text, watermark

Parameters for this clip

ParameterValueWhy
Duration10sLong enough for the turn and smile to unfold without rushing
Resolution1080pA close-up portrait — skin and eye detail matter
Aspect ratio9:16A single subject, close-up — vertical works better
Cameraslow push-inSupports the calm action, adds engagement
camera_fixedoffWe want a soft dolly-in, so we don't lock the camera
generate_audiooptionalUsually not needed for a silent portrait
The one-lever rule
Didn't like the result? Change one thing and regenerate. Turn too weak — strengthen the verb ("turns fully toward the camera"). Camera jittering — slow the push-in or turn on camera_fixed. This way you learn which lever does what, instead of changing everything at once and guessing.
Key takeaway
The model already sees the starting frame. Your prompt is a motion script: action verb → camera to match it → pacing → light changes → the arc of "beginning and how it all settles." Describe the motion, not the picture — and your first clip will come alive.
[Image 1]push-in10s / 1080pnegative_promptone lever

Try it yourself in ZenCreator

Generate your first clip — everything this course teaches is available in the ZenCreator video tools.

Start Creating
PreviousNext lesson