AI Video Prompting — Part 1: Camera & Actions

Series: Mastering Video Prompting (3 parts)
Author: François Amat — Date: March 16, 2026
Platform: ElevenLabs Image & Video
Model: Seedance 1.5 Pro (keyframe-to-keyframe)

---

Introduction — Video Prompting, a New Language

AI video generation has taken a spectacular leap forward. Today, platforms like ElevenLabs, Runway, and Google Veo let you create video clips from simple text descriptions. But unlike image generation, video prompting requires a fine understanding of three simultaneous dimensions: camera movement, in-scene action, and visual atmosphere.

AI video generation isn't just about writing "make me a video." Video prompting is a directing language: every word in the prompt controls a specific aspect of the result.

This 3-part guide is the result of systematically testing 34 prompts on ElevenLabs with Seedance 1.5 Pro. For each technique:

• The keyframes for start and end (Start Frame / End Frame)
• The exact motion prompt used
• The resulting video generated by the model

📖 In this series:

• Part 1 (this article) — Camera Movements & Actions/Transitions
• Part 2 — Atmosphere, Speed & Composite Combinations
• Part 3 — Golden Rules, Compatibility & Final Edits

---

How Does Keyframe-to-Keyframe Work?

┌──────────────┐     ┌──────────────────┐     ┌──────────────┐
│  START FRAME │ ──▶ │  MOTION PROMPT   │ ──▶ │  END FRAME   │
│  (Image A)   │     │  (Instructions)  │     │  (Image B)   │
└──────────────┘     └──────────────────┘     └──────────────┘

The model interpolates between two images following the motion prompt instructions. A prompt always follows the same structure:

[CAMERA MOVEMENT], [SUBJECT ACTION], [ATMOSPHERE], [SPEED]

The Test Scene

To isolate the effect of each command, all tests use the same base scene: a mountain lake at dawn with mist, pine trees and golden light. Only the tested parameter changes each time.

---

Camera Movements 🎥

Camera terms control the viewpoint displacement between the two keyframes. This is the foundation of every video clip.

Golden rule: One camera movement per clip. Combining two = guaranteed artifacts.

☝️ Overview of camera movements on the lake scene

---

1. slow push in (Slow Zoom In)

Effect: The camera gradually advances toward the subject. The frame tightens, the background moves away through parallax. Creates a feeling of intimacy and focus.

When to use: Emotional transitions, close-up on a detail, intimate revelation moments.

Motion Prompt:

slow push in, mist gently moves across the lake surface, smooth motion, cinematic

---

2. slow pull back (Slow Zoom Out)

Effect: The camera slowly pulls back, progressively revealing the spatial context around the subject. Creates a "reveal" effect — we discover where the scene takes place.

When to use: End of sequence, spatial reveal, transition from detail to full landscape.

Motion Prompt:

slow pull back, mist gently moves across the lake surface, smooth motion, cinematic

---

3. static camera (Fixed Camera)

Effect: No camera movement. The frame is locked. Only scene elements change — water, mist, light. Isolates the change.

When to use: When the viewer needs to focus on what moves in the scene. Ideal for morphing and light transitions.

Motion Prompt:

static camera, mist gently moves across the lake surface, smooth motion, cinematic

---

4. dolly left (Left Tracking Shot)

Effect: The camera slides horizontally to the left, creating parallax scrolling: closer objects move faster than distant ones. Adds depth.

When to use: Panoramas, transitions between adjacent elements, lateral subject tracking.

Motion Prompt:

dolly left, mist gently moves across the lake surface, smooth motion, cinematic

---

5. dolly right (Right Tracking Shot)

Effect: Same mechanics as dolly left, opposite direction. The camera slides to the right.

Motion Prompt:

dolly right, mist gently moves across the lake surface, smooth motion, cinematic

---

6. crane up (Upward Movement)

Effect: The camera rises vertically — from ground level to an overhead view. The scene's scale is progressively revealed.

When to use: Dramatic reveals, showing a location's scale, transitions to an overview.

Important: Keyframes must reflect both altitudes — Start at ground level, End aerial. Without this consistency, the model doesn't understand the movement.

Motion Prompt:

crane up, mist gently moves across the lake surface, smooth motion, cinematic

---

7. crane down (Downward Movement)

Effect: Reverse of crane up. The camera descends from an aerial view to the ground. Landing effect.

When to use: Introduction to a location, focus on a ground-level detail, descending transition.

Motion Prompt:

crane down, mist gently moves across the lake surface, smooth motion, cinematic

---

8. orbit around (Orbital Rotation)

Effect: The camera describes an arc around the subject. The subject stays centered while the angle changes. Very cinematic — reminiscent of signature action movie shots.

When to use: Showcasing a character or object, slow "bullet time" effect, presentations.

Tip: For best results, Start/End frames should show the subject from slightly different angles (15-30°).

Motion Prompt:

orbit around, mist gently moves across the lake surface, smooth motion, cinematic

---

9. handheld (Handheld Camera)

Effect: Micro-tremors in the framing. Simulates the natural instability of a handheld camera. Documentary style, immersive.

When to use: Realism, tension/urgency scenes, reportage style. It's a modifier — it combines with other movements.

Motion Prompt:

handheld, mist gently moves across the lake surface, smooth motion, cinematic

---

10. smooth tracking shot (Smooth Tracking)

Effect: The camera smoothly follows a moving subject, maintaining its position in the frame. The scenery scrolls in the background.

When to use: Following a walking character, smooth action sequences, travel shots.

Motion Prompt:

smooth tracking shot, mist gently moves across the lake surface, smooth motion, cinematic

---

Summary — Camera Movements

Tips:

• One camera movement per clip to avoid artifacts
• orbit around requires start/end frames from different angles
• handheld is a modifier — add it to another movement

---

Actions and Transitions 🔄

Action terms describe what changes in the scene between the two keyframes. This is the narrative core of each clip. Each action was tested with a fixed camera (static camera) to isolate its effect.

☝️ Overview of actions and transitions on the lake scene

---

1. gradually transforms into (Morphing)

Effect: Image A deforms pixel by pixel to become image B. An organic transition between two visual states. The model finds correspondence points and progressively deforms.

When to use: Season changes, aging, material transformation, magical effects.

Key tip: Morphing works MUCH better when both frames share the same composition (same angle, same subject). Only the state changes.

Motion Prompt:

static camera, the lake scene gradually transforms into, smooth motion, cinematic

---

2. particles scatter (Particle Dispersion)

Effect: Particles (light, dust, sparks) appear and scatter in all directions. Dissolution effect, released energy.

When to use: Soft explosions, poetic disintegration, snow/pollen effects, magical moments.

Motion Prompt:

static camera, the lake scene particles scatter, smooth motion, cinematic

---

3. particles converge (Particle Convergence)

Effect: Reverse of scatter — dispersed particles converge to a central point. Assembly effect, materialization.

When to use: Object appearance, materialization, energy concentration, shape formation.

Motion Prompt:

static camera, the lake scene particles converge, smooth motion, cinematic

---

4. light intensifies (Light Rise)

Effect: Scene brightness increases progressively. Transition from darkness to light. Sunrise, illumination effect.

When to use: Sunrise, luminous reveal, climax moments, night → day transition.

Note: Keyframes are adjusted — Start is dark (pre-dawn) and End is bright. This consistency is essential.

Motion Prompt:

static camera, the lake scene light intensifies, smooth motion, cinematic

---

5. light fades (Light Decline)

Effect: Lighting progressively decreases. The scene darkens, shadows lengthen. Sunset effect.

When to use: Sunset, end of sequence, entering a dark place, descent into mystery.

Motion Prompt:

static camera, the lake scene light fades, smooth motion, cinematic

---

6. reveals (Reveal)

Effect: A hidden element appears progressively. The mist lifts, the shadow recedes — and something is unveiled. Creates narrative suspense.

When to use: Mystery reveal, character introduction, location discovery.

Motion Prompt:

static camera, the lake scene reveals, smooth motion, cinematic

---

7. dissolves into (Cross Dissolve)

Effect: Image A becomes semi-transparent while image B emerges underneath. The classic cinematic cross dissolve, reinterpreted by AI.

When to use: Time transitions, dream passages, location changes. The safest for connecting two very different images.

Motion Prompt:

static camera, the lake scene dissolves into, smooth motion, cinematic

---

8. emerges from (Emergence)

Effect: A subject gradually comes out of an environment — emerging from water, mist, shadow. Very dramatic effect.

When to use: Character appearances, births, crisis resolution, rebirth moments.

Motion Prompt:

static camera, the lake scene emerges from, smooth motion, cinematic

---

9. ripple effect (Ripple)

Effect: Concentric ripples distort the image, like a stone thrown into a water mirror.

When to use: Aquatic transitions, emotional shock effects, reality distortions, dreamlike moments.

Motion Prompt:

static camera, the lake scene ripple effect, smooth motion, cinematic

---

10. glitch transition (Digital Distortion)

Effect: Digital artifacts — distortion bands, broken pixels, RGB shifts, fragmentation. The clip visually "glitches." Very stylized.

When to use: Cyberpunk/tech style, music clips, brutal transitions, narrative ruptures.

⚠️ Warning: The riskiest term — the model may over-interpret and produce chaotic results.

Motion Prompt:

static camera, the lake scene glitch transition, smooth motion, cinematic

---

Summary — Actions and Transitions

Tips:

• gradually transforms into works best when frames share the same composition
• dissolves into is the safest for connecting very different images
• glitch transition is the riskiest — the model may over-interpret

---

Coming Up...

In Part 2, we explore atmosphere and mood terms (cinematic lighting, dreamlike, volumetric fog...), speed controls, and especially composite combinations — when you mix camera, action, mood and rhythm in a single prompt.

---

Platform: ElevenLabs Image & Video — Model: Seedance 1.5 Pro — Scene: Mountain lake at dawn