CRITICALFailure Reference

Luma Dream Machine Lip Sync Failure — Refund Guide

Technical Classification

Phoneme-Viseme Mapping Failure

Phoneme-Viseme Mapping Failure on Luma Dream Machine occurs when the lip-shape predictor outputs visemes uncorrelated with the input or generated audio. Mouth opens on consonants, closes on vowels, lips fail to round on "oh" sounds, and overall lip-movement timing drifts from audio. Distinct from prompt-adherence failures because the audio track itself is correct — only the visual mouth-shape sequence is wrong.

How to identify this failure

✕Mouth shape uncorrelated with audio phoneme
✕Closed mouth during open vowels
✕Lip rounding wrong for "oo" / "oh"
✕Mouth motion timing drifts from audio across clip
✕Generic chewing motion replaces phoneme-specific visemes

Real generation examples

Prompt used

"Woman saying \"good morning\" directly to camera, professional lighting"

Failure observed @ 0:00 → 0:02

Mouth stays nearly closed through entire phrase; brief opening at 0:01 not aligned with phonemes

Prompt used

"News anchor delivering a single sentence, lip-synced to audio"

Failure observed @ full duration

Mouth motion is generic chewing pattern; no correlation with audio sentence structure

Documentation strength

If you need to escalate

HIGH — Luma Dream Machine markets dialogue capability; lip-sync failure on paid output is treatable as feature defect.

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 Phoneme-Viseme Mapping 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 Phoneme-Viseme Mapping 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 "Phoneme-Viseme Mapping 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

Does Luma support response lip-sync failures?

Yes — Luma support recognises lip-sync as a marketed Dream Machine capability. Cite "Phoneme-Viseme Mapping Failure", attach the audio + visual stills.

Why does Luma lip-sync fail?

The viseme-prediction network is conditioned on audio embeddings. Under fast speech, accented audio, or non-English phonemes, the embedding-to-viseme mapping degrades, producing generic mouth motion.

Which Luma dialogue prompts are highest risk?

Fast speech, non-English audio, accented dialogue, multi-speaker scenes, and any prompt where lip-sync is narratively load-bearing. AVA flags lip-sync prompts before generation.

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