Runway Act-Three 2026: AI Motion Capture Ads Guide

You’re pitching UGC-style ad variants at the volume Meta and TikTok now demand, and every one still needs a human on camera. A single 30-second spot means booking talent, scheduling a shoot, paying usage fees that expire in 90 days, and then re-shooting the whole thing when the hook underperforms. Meanwhile the AI video you tried last year still produces the tell: dead eyes, lip-sync drifting by two frames, a character whose face subtly re-rolls between shot 3 and shot 4. Act-Three changes the math — but only if you know where performance transfer holds and where it visibly collapses.

This is for intermediate creators, performance marketers, and small production teams already shipping paid social — people who understand shot lists, aspect ratios, and why a hook dies at 1.5 seconds. You should be comfortable inside a browser-based generation tool and able to read a credit balance without hand-holding. Out of scope: beginner “what is AI video” material, long-form narrative filmmaking, custom model training, code or API integration, and any promise that this replaces a director’s judgment.

Honest assessment: Act-Three is genuinely strong at transferring a driving performance’s timing, head motion, and gross expression onto a target character — that’s the part that used to be impossible. It is unreliable at sustained eye-line under off-axis capture, at plosive-heavy lip-sync, at hands, and at holding one identity across a multi-shot spot without deliberate intervention. Human review is non-negotiable at three points: reference image approval before you spend credits, a frame-level pass on lip-sync and eyes before upscaling, and a policy/disclosure check before anything touches an ad account. Nobody should be publishing unreviewed output — likeness and platform policy exposure is real.

What This Guide Covers

• Understand what performance transfer actually changed about ad production economics in 2026 — and what it didn’t
• Read the transfer pipeline clearly enough to diagnose a bad output instead of re-rolling blind
• Run real cost math across Unlimited, Standard, and Pro so you know your true spend per finished second before you commit
• Prep reference images that survive transfer intact, using the criteria that separate a clean result from an uncanny one
• Set up webcam or phone capture for driving performances without a studio, lighting rig, or crew
• Walk one complete first shot end to end, with the decision points called out where most people go wrong
• Spot and correct the three failure modes — lip-sync, eye-line, micro-expression — that make output read as AI
• Hold a single character consistent across an entire multi-shot spot instead of getting a new face every clip
• Combine Act-Three with Gen-4.5 and Aleph so transferred performance lands inside finished, gradeable shots
• Hit the upscaling, frame rate, and deliverable specs Meta and TikTok actually want, not the ones you’re guessing at
• Choose deliberately between Act-Three, Hedra Character-3, Kling Motion Brush, and Veo based on what each is genuinely better at
• Follow a full 30-second UGC-style ad from concept through delivery, including the parts that don’t work the first time
• Rejection-proof your creative against ad platform policy and AI disclosure rules with a concrete pre-flight fix list
• Handle licensing and likeness releases properly, and compare cost per finished minute against live talent with real numbers

Instant online access the moment checkout completes — the full guide is available immediately, no waiting, no email sequence, no upsell.