Runway Gen-3

Overview

Runway Gen-3 Alpha is the leading commercial AI video generation model. Known for consistent motion, temporal coherence, and high-quality cinematic output up to 10 seconds per generation.

Key Strengths

• Temporal consistency: Subjects maintain identity across frames
• Camera control: Specify camera movements (pan, tilt, dolly, orbit)
• Motion quality: Natural, physics-aware motion
• Image-to-video: Animate still images with specified motion
• Motion Brush: Paint motion onto specific areas of an image

Prompting Tips

Camera Movement Vocabulary

"Slow dolly in on..." — camera moves toward subject
"Tracking shot following..." — camera follows moving subject
"Static shot of..." — no camera movement
"Orbit around..." — 360° rotation around subject
"Crane up revealing..." — camera rises vertically
"Push in on..." — forward camera movement

Effective Video Prompts

GOOD: "Slow dolly forward through a misty forest at dawn, rays of light filtering through tall pine trees, a deer standing in a clearing turns its head toward camera, cinematic, shallow depth of field"

BAD: "A pretty forest" (too vague, no motion described)

Key Guidelines

1. Describe ONE motion or action (don't overload)
2. Specify the camera movement explicitly
3. Include lighting and atmosphere
4. Keep prompts concise (50-100 words)
5. Use "cinematic" for film-quality output

Best Use Cases

Use Case	Quality
Cinematic establishing shots	Excellent
Product showcase rotations	Excellent
Nature and landscape motion	Very Good
Character animation	Good
Complex multi-person scenes	Fair
Long sequences (>10s)	Requires stitching

Image-to-Video

Upload a still image and describe the motion:

Image: [photo of a woman at a café]
Prompt: "She slowly looks up from her book and smiles, steam rising from the coffee cup, warm afternoon light"

Pricing

• Standard: $0.25 per second of video
• 10-second clip: $2.50
• Generation time: ~30-90 seconds