Runway Gen-2 Tutorial: Turning Text Into Stunning Video Clips

Runway Gen-2 Tutorial: Turning Text Into Stunning Video Clips

The first time I generated a video of ocean waves crashing against a cliff at sunset — complete with spray, motion, and cinematic camera movement — from a single text prompt in under 30 seconds, I stopped and watched it three times.

Not because it was perfect. It wasn't. But because it existed at all, from text, in half a minute. That shift in what's possible for a solo video creator without a film crew is significant.

This Runway Gen-2 tutorial covers everything you need to start generating AI video clips: how the tool works, the prompting techniques that produce the best results, and how to integrate AI video into a real production workflow.

Getting Started With Runway

Creating your account: Go to runwayml.com and sign up. New accounts receive 125 free credits — enough for several test generations to evaluate the tool before committing to a subscription.

Navigating the interface:

• Generate — the main creation area (text-to-video, image-to-video)
• Assets — your generated videos and uploaded media
• Editor — Runway's built-in video editor (separate from generation)
• Models — access to Gen-2, Gen-3, and other Runway tools

For this tutorial, we'll focus on the Generate section.

Text-to-Video: Your First Generation

Step 1: Select the model In the Generate section, select "Gen-3 Alpha" for the best results (Gen-2 is still available but Gen-3 quality is significantly better for most use cases).

Step 2: Write your prompt This is where most beginners go wrong — they describe a concept rather than a visual scene. Effective video prompts describe:

• What you can see in the frame
• Camera movement or perspective
• Lighting and atmosphere
• Subject motion

Weak prompt: "a peaceful forest"

Strong prompt: "Slow dolly forward through a misty old-growth forest, morning light filtering through tall redwood trees, soft fog at ground level, silence, photorealistic, cinematic"

Step 3: Set parameters

• Duration: 4–10 seconds (Gen-3 Alpha)
• Aspect ratio: 16:9 for standard video, 9:16 for social/vertical
• Resolution: 720p (faster) or 1080p (higher quality, more credits)

Step 4: Generate Click Generate and wait 20–60 seconds. You'll receive 1 video output. If you want variations, generate again with the same prompt — each generation is different.

Prompting Techniques for Better Results

Camera Movement Vocabulary

Including camera movement language dramatically improves cinematic quality:

• "slow dolly forward" / "dolly back" — camera moves toward/away from subject
• "pan left" / "pan right" — camera rotates horizontally
• "tilt up" / "tilt down" — camera rotates vertically
• "crane shot" — camera moves up and back to reveal environment
• "handheld, slightly shaky" — adds naturalistic movement
• "static locked camera" — no camera movement, subject moves
• "orbital" — camera circles around subject

Lighting Descriptors That Work

• "golden hour" / "magic hour" — warm directional sunlight
• "blue hour" — just before/after sunset, cool tones
• "harsh midday sun" — high contrast, strong shadows
• "overcast, diffused light" — soft, shadowless
• "neon-lit" / "nighttime urban" — artificial colored light
• "candlelit" / "firelight" — warm, flickering
• "studio lighting" — controlled, professional

Motion and Atmosphere

• "time-lapse clouds rushing" — speeds up naturally slow motion
• "slow motion" — the opposite, for dramatic effect
• "steam rising" / "dust particles floating" — atmospheric micro-motion
• "leaves rustling" — adds life to static scenes
• "water reflection rippling" — natural motion in reflections

Image-to-Video: The More Reliable Approach

For most professional use cases, Image to Video produces more usable and predictable results than Text to Video.

Why Image to Video is better for production:

• You control the initial composition precisely (generate the perfect image first with Midjourney or DALL-E 3)
• The video motion applies to a known composition rather than generating everything from scratch
• Results are more consistent and predictable

Workflow:

1. Generate a high-quality still image using Midjourney or DALL-E 3
2. Upload to Runway's Image to Video
3. Describe the motion you want applied: "gentle camera push forward," "subtle motion in the leaves," "zoom slowly out to reveal the full scene"
4. Generate

Example: I created a fantasy landscape image in Midjourney, then used Runway Image to Video to animate it with a slow aerial camera push forward. The result was a cinematic establishing shot that looked like footage from a film production.

Practical Use Cases and Example Prompts

Social Media B-Roll

Prompt: "Aerial view of a city at night, slow tilt down from skyline to street level, neon reflections on wet pavement, cinematic, shallow depth of field on street below"

Use: Background for social media content, YouTube intros, podcast video backgrounds

Nature and Landscape

Prompt: "Ocean waves crashing against rocky coastline, morning fog, slow motion spray, wide establishing shot, golden hour light, photorealistic"

Use: Documentary-style B-roll, environmental content, meditation video content

Technology/Abstract

Prompt: "Data streams flowing through glowing fiber optic cables, extreme close up, rapid motion, blue and white light, abstract technology visualization"

Use: Tech channel content, SaaS product videos, explainer video backgrounds

Urban/Architecture

Prompt: "Pedestrians crossing a busy intersection in Tokyo at night, slow motion, neon signs reflecting in puddles, crane shot from above, cinematic"

Use: Travel content, brand videos, establishing shots

Extending Short Clips for Longer Sequences

Gen-2 and Gen-3 generate short clips. For longer video sequences, use the Extend feature:

1. Generate your initial clip
2. Click "Extend" to continue generating from the last frame
3. Add a new text prompt describing what should happen next
4. Chain 3–5 extensions for longer sequences

This approach maintains visual continuity between clips — the camera and scene stay consistent across extensions.

Integrating Runway Into Your Editing Workflow

Runway video clips are most powerful as B-roll in edited videos, not as standalone pieces. My workflow:

1. Script and record main video (or use AI voiceover)
2. Identify B-roll needs from the script
3. Generate Runway clips for abstract or impossible-to-film elements (aerial views, historical settings, abstract concepts)
4. Supplement with stock footage (Pexels) for simpler requirements
5. Assemble in CapCut or Premiere Pro

The rule: use Runway for content that genuinely can't be filmed practically or cost-effectively. Use stock footage for standard coverage. The combination creates a visually varied video that doesn't look entirely AI-generated.

Frequently Asked Questions

What is Runway Gen-2?

Runway's AI video generation model for creating short video clips from text or images. Continuously improved since 2023; largely superseded by Gen-3 Alpha for quality but still available.

Is Runway Gen-2 free?

125 free credits for new accounts. Paid plans from $15/month (625 credits). A 4-second video costs approximately 20–40 credits.

How long are Runway videos?

4–10 seconds per generation (Gen-3 Alpha). Clips can be extended by chaining generations. Multiple clips are assembled in video editing software for longer sequences.

What's the difference between Gen-2 and Gen-3?

Gen-3 Alpha has better motion quality, longer generation (up to 10 seconds), and more precise camera control. Gen-3 is recommended for most new work.

Can Runway generate video from an image?

Yes — Image to Video is one of the most useful features, producing more predictable results than text-to-video by letting you control initial composition precisely.

Final Thoughts

Runway is the most accessible professional-grade AI video tool available. The text-to-video and image-to-video features, once you develop prompting fluency, produce clips that integrate naturally into video productions.

The 125 free credits are enough to develop real intuition about what the tool produces from different prompts. Spend them on variety — different scene types, different camera movements — to build a sense of what the model does well.

For a broader video production workflow using AI tools, our guide on building a faceless YouTube channel shows how Runway fits into a complete production stack. And for a comparison of Runway against Stable Diffusion's video capabilities, see our Runway ML vs Stable Diffusion deep-dive.