CRITICALFailure Reference

Runway ML Face Distortion — Facial Landmark Regression Failures

Technical Classification

Facial Landmark Regression Failure

Facial Landmark Regression Failure occurs when Runway's face generation pipeline produces topologically invalid facial geometry or fails to maintain consistent facial identity across frames. The failure manifests as asymmetric landmark placement, morphological drift (the face gradually changes shape over the clip), or non-Euclidean proportions where nose width, eye spacing, or mouth position falls outside anatomical norms. Face distortion is especially severe in close-up shots and when the subject moves or turns their head.

How to identify this failure

✕Eyes misaligned — one significantly higher than the other
✕Face gradually morphing or melting across the clip duration
✕Nose or mouth in anatomically impossible positions
✕Facial proportions changing between frames
✕Uncanny valley distortion in close-up shots

Real generation examples

Prompt used

"A news anchor sitting at a desk, looking directly into camera"

Failure observed @ 0.4s

Left-eye displacement ~15% of face width, morphological drift at 0.3s intervals

Documentation strength

If you need to escalate

VERY HIGH — Face distortion is visually unmistakable and clearly makes the output commercially unusable.

AVA is a pre-purchase prevention tool, not a post-purchase recovery tool. Platforms generally do not guarantee credit refunds for output-quality failures; goodwill credits are at each platform's discretion. The strength rating reflects how well-formed your support ticket can be, not a promised outcome.

Prevention + documentation steps

01
Score your prompt before you generate
Run your prompt through AVA's pre-flight scoring against the Facial Landmark Regression Failure pattern. Green light = generate. Yellow/red = rewrite using the suggested fix before you commit credits.
02
Capture Generation ID + timestamp if it failed anyway
Find the Generation ID in the URL or share link. Note the exact time when the Facial Landmark Regression Failure first appears (e.g. "failure first visible at 1.2s"). Timestamped evidence is significantly stronger than a general complaint.
03
Use the correct technical term in your support ticket
Describe this failure as "Facial Landmark Regression Failure". This term maps to a recognised internal workflow in the support system and routes the ticket to the right team.
04
Submit via the correct support channel
Runway has no direct email intake. Pro+ plan: open the in-app AI Assistant (help widget bottom-right of app.runwayml.com), describe the failure with the technical term, attach evidence. Free/Standard plan: human support isn't available — your channel is Discord #community-help with @On Call - Moderators.

Frequently asked questions

Will Runway support escalate for face distortion?

Yes. Face distortion is classified as a critical failure mode by Runway support. Reference "Facial Landmark Regression Failure" and include the timestamp where distortion first appears along with your Generation ID.

Why does Runway distort faces?

The identity-preserving latent space traversal breaks down when the subject moves or the camera angle changes. The model must interpolate facial geometry frame-by-frame, and drift accumulates over time.

Are Runway face failures worse in certain conditions?

Yes. Close-up shots, head movement, multiple subjects, and longer clips all increase face distortion risk. If you need a clean talking head, keep it under 4 seconds with minimal movement.

Catch it before you generate

AVA scores this failure mode against your prompt in real time

Free Chrome extension. Analyzes your prompt as you type, flags failure-prone patterns specific to this model, and tells you what to rewrite — before you commit credits to a generation that will fail.

Install free All 105 failure modes

AVA Pro · founders' round

$50 for 6 months of unlimited scoring across all failure modes + personal failure-history dashboard. Locks in $13/mo grandfathered after.

Claim $50 founders