Gen-4 is Runway's current production model, and Gen-4 Turbo is its faster variant. Both are good enough for paid social if you know how to prompt them. This tutorial covers the exact steps we follow when producing DTC ad clips with Gen-4.
Step 1: Choose Gen-4 Standard or Gen-4 Turbo
| Gen-4 | Gen-4 Turbo | |
|---|---|---|
| Resolution | Up to 4K upscale | Up to 1080p native |
| Max duration | ~16s per generation | ~10s per generation |
| Motion quality | Higher fidelity, better physics | Slightly softer on fast motion |
| Speed | ~90s per clip | ~30s per clip |
| Best for | Hero ads, product close-ups | Iterating concepts, batch variations |
Start with Turbo for testing concepts. Switch to standard Gen-4 for final renders.
Step 2: Prep Your Input Image
Gen-4 performs best with image-to-video rather than pure text-to-video for ad work. Generate your starting frame in FLUX 1.1 Pro Ultra or Midjourney, then feed it into Gen-4.
Key rules for input images:
- Isolate the product on a clean background or lifestyle setting. Busy compositions confuse the model.
- Match your final aspect ratio (9:16 for Reels/TikTok, 1:1 for feed, 16:9 for YouTube pre-roll). Cropping after generation wastes resolution.
- Light the product the way you want it in motion. Gen-4 preserves the lighting direction from your input frame with high consistency.
Step 3: Write the Motion Prompt
Gen-4 responds well to short, specific motion descriptions. Skip the poetic language.
Bad prompt: "A beautiful cinematic shot of the product floating gracefully through a dreamy ethereal space."
Good prompt: "Camera slowly pushes in toward the bottle. Condensation droplets roll down the glass. Soft bokeh background shifts slightly."
Structure every prompt with three elements:
- Camera move (push in, orbit, static, pan left)
- Subject action (liquid pours, hand picks up jar, fabric drapes)
- Environment detail (background blurs, light flares from right, shadows shift)
Step 4: Use Camera Control Presets
Gen-4 offers explicit camera motion controls. Use them instead of describing camera movement in your text prompt, as the controls override prompt language anyway.
For ads, these three moves cover 80% of needs:
- Push in (slow): Hero product reveals, building desire.
- Orbit (quarter turn): Showing product dimensionality, works well for packaging.
- Static + subject motion: Let the product or model move while the camera stays locked. Feels most like a real ad shoot.
Step 5: Generate, Review, Extend
Generate 3 to 4 variations of each shot. Gen-4 is not deterministic, so you will get meaningfully different motion interpretations. Pick the best, then extend if you need more duration.
When extending clips:
- Keep extensions to one round (adding ~5 to 8 seconds). Multiple extensions degrade quality.
- Write a new prompt for the extension that describes where the motion should go next. Do not reuse the original prompt.
Step 6: Export and Post-Process
Export at the highest available resolution. Bring the clip into your editor (CapCut, Premiere, DaVinci) for:
- Color grading to match brand palette
- Adding a CTA overlay and logo
- Sound design (Gen-4 does not generate audio natively, unlike Veo 3)
- Trimming the first and last 5 to 10 frames, which sometimes contain generation artifacts
At Adsome, we typically generate 8 to 12 raw clips per ad concept in Gen-4, then cut the best 3 to 4 seconds from each into a single 15 to 30 second edit. This gives you the consistency of AI generation with the pacing of a real ad.
Verdict
Gen-4 is currently one of the top two models for product ad clips (alongside Kling 3.0). Its main advantage is camera control precision. Its main weakness is audio, which you will need to add in post. For DTC brands running Meta and TikTok ads, Gen-4 with a FLUX-generated input frame is a reliable, repeatable pipeline.
