Runway Gen 4.5 review: realistic AI video physics (full demo)

Runway Gen 4.5 dropped with a bold promise: realistic physics, less morphing, smoother motion, and a generation engine that finally understands real-world cause and effect. I put it through seven progressively harder tests and compared the output to the AI Media Machine side-by-side. Here's the truth.

Test 1: The simple shot (elderly woman walking)

Clean. Believable. One subtle flaw — but the average viewer wouldn't notice. Off to a strong start.

Test 2: Complex movement (dancer)

This is where Gen 4.5 stumbled. Limbs morphed mid-rotation, fingers blended into the dress, the camera held too still while the body asked for a follow. Complex motion is still the hardest thing in AI video.

Tests 3-4: Extreme action (parkour) vs. cartoon style (samurai)

Realism collapsed at the parkour test — the model fought itself trying to keep limbs attached to bodies during impact. Then I switched to a stylized samurai shot and it sang. The lesson: Gen 4.5 punches harder when you give it permission to be stylized.

Tests 5-7: The good, the fast, and the limits

The homepage example test produced the cleanest shot of the day. Fast-motion kitchen and motorcycle shots came in surprisingly stable. The cat-and-can advanced physics test exposed the wall — Gen 4.5 sees physics but doesn't simulate them. It approximates.

The pricing is genuinely confusing

Runway's pricing page is the kind of thing you need a spreadsheet for. Multiple plans, credit costs that change by feature, and a Pro tier that still throttles certain models. Read carefully before subscribing.

The big problem: no image-to-video

Here's what nobody covers. Gen 4.5 doesn't ship with image-to-video at launch. For a creator workflow that depends on character and product consistency — feeding the same hero image into multiple shots — that's a dealbreaker. You're forced back to Gen 4 or to another tool entirely.

Side-by-side vs. AI Media Machine

I ran the same prompts through the AI Media Machine with image-to-video on. Same character, same product, same brand colors across all shots. Gen 4.5 alone can't deliver that consistency yet, and that's the whole game for ad creative.

Six brutal comparisons (cost, resolution, audio, lip-sync, image-to-video, batch)

The comparison table in the video lays out cost-per-clip, native resolution, audio support, lip-sync quality, image-to-video availability, and batch generation. AIMM wins five of six categories at a fraction of the price.

The final verdict — who is Gen 4.5 actually for?

Filmmakers who want stylized B-roll and have an editor on standby. Not for ad creators, not for product marketers, not for anyone who needs character consistency across 6 shots in 6 minutes.

The two paths forward

If you want the all-in-one stack with image-to-video, lip-sync, and 12 connected apps, try the AI Media Machine for $1. If you want us to build your AI video system for you, book a free strategy call.