AI Video — Runway, Sora, and smart editing

Section 10.5.4 : AI Video — Runway, Sora, and smart editing

🎯 Learning objective

Explore AI video generation and editing tools — Runway Gen-3, Sora 2, Google Veo 2, and editing tools — to create professional video content without a production team. You'll understand the capabilities and limitations of each tool, and master the complete AI video production workflow.

The AI video revolution

AI video generation has experienced explosive acceleration in 2024-2026. What previously required a production team (director, camera operator, editor, studio), a budget of several thousand dollars, and weeks of work can now be created in minutes by a single person with a monthly subscription.

This doesn't replace cinematic productions — but for marketing content, social media, e-learning, and presentations, AI video is a game-changer.

Video generation tools: complete comparison

Sora 2 in detail (2025-2026 leader)

Sora 2 by OpenAI is the most advanced video generation model currently available:

Capabilities:

• →Generation of realistic videos up to 60 seconds in cinematic quality (1080p)
• →Physics understanding (gravity, reflections, consistent shadows, fluids)
• →Excellent temporal consistency (objects don't "mutate" between frames)
• →Understanding of 3D spatial relationships
• →Integrated Storyboard: plan multi-shot sequences with per-scene descriptions
• →Remix: upload an existing video and transform it (change style, environment, season)
• →Available on iOS and Android since late 2025

Current limitations:

• →Hands and fingers sometimes inconsistent (like with images)
• →Complex body movements (dance, sports) sometimes unnatural
• →On-screen text not guaranteed
• →Generation time: 1-5 minutes for a 10s video

Effective video prompts for Sora 2:

Recommended structure:
[Scene description] + [Camera movement] + 
[Action/movement] + [Style/mood] + [Lighting] + [Duration]

Example — Product advertisement:

Close-up shot of a premium coffee mug on a wooden desk. 
Steam rises gently from the dark coffee. Camera slowly orbits 
around the mug, revealing a minimalist logo. Morning golden-hour 
light comes through a window on the left. Soft bokeh background 
of a modern office. Cinematic quality, warm color grade. 10 seconds.

Example — Presentation transition:

Aerial drone shot of a modern city transitioning from day to 
night through a timelapse effect. Buildings light up progressively. 
Camera smoothly descends towards street level. Cinematic, 
commercial quality. 8 seconds.

Runway Gen-3 Alpha in detail

Runway is the go-to tool for precise motion control:

Motion Brush (unique feature): Upload an image → paint the areas you want to animate → indicate the direction and intensity of movement. The rest of the image stays static. Perfect for: animating a product photo, making hair move, creating a parallax effect.

Generation modes:

• →Text-to-Video: describe the scene → video generated
• →Image-to-Video: static image → animation (most used, as you control the starting point)
• →Video-to-Video: transform an existing video's style (e.g., real video → anime style)

Camera Controls (Gen-3 Alpha Turbo): Precisely control camera movement:

• →Progressive zoom in/out
• →Horizontal/vertical pan
• →Rotation / orbit
• →Movement combinations

The complete AI video production workflow

Here's the step-by-step process for creating a professional video with AI, from script to publication:

Total time: 2-3 hours for a 2-5 minute professional quality video. Comparison: an equivalent traditional video production: 2-5 days + $2,000-10,000.

AI video editing: post-production tools

Descript in detail (revolutionary): Import your video → Descript automatically transcribes it → you edit the video by editing the text. Delete a word from the transcript → the corresponding video passage is automatically cut. Add text → Descript generates the corresponding voice (with your cloned voice). It's the most intuitive video editor in existence.

Practical exercise: create a 30-second ad video

1. →Script (ChatGPT):

Write a script for a 30-second ad video 
for [product/service]. Structure: hook (3s) → problem (5s) 
→ solution (10s) → visual demo (7s) → CTA (5s). 
Include visual descriptions for each shot.

2. →

   Key images (DALL-E 3/Midjourney): generate 5 images matching the 5 shots of the script

3. →

   Animation (Runway or Sora 2): animate each image into a 5-10 second clip

4. →

   Voiceover (ElevenLabs): generate the script narration

5. →

   Assembly (CapCut): combine clips, add voiceover, music, and subtitles