CRITICALFailure Reference

Luma Dream Machine Hand Artifact — Pre-Generation Risk Reference

Technical Classification

Hand-Anatomy Topology Failure

Luma Dream Machine produces Hand-Anatomy Topology Failures at a rate comparable to its closest competitors. The Ray-2 model improved face coherence significantly but hand topology remains under-constrained — the model generates 5 fingers most of the time but loses count on close-ups, motion sequences, and any clip where hands are the visual subject. The failure compounds over the clip length: hands look fine at frame 1 and topologically wrong by frame 60.

How to identify this failure

  • Finger count drifting between 4, 5, and 6 across frames
  • Knuckles bending laterally instead of along anatomical axis
  • Thumb merging into index finger
  • Hand orientation flipping between palm and back
  • Hand geometry collapsing during motion

Real generation examples

Prompt used

"Chef chopping vegetables, close-up of hands and knife"

Failure observed @ 0:02

Left hand showed 4 fingers at 0:02, 6 fingers at 0:04; knife handle merged with thumb

Prompt used

"Musician's hands playing acoustic guitar"

Failure observed @ 0:03

Strumming hand lost finger separation at 0:03; 6 fingers visible on fretting hand at 0:05

Documentation strength

If you need to escalate

HIGH — Luma support refunds documented hand-anatomy failures under their per-generation credit policy. Provide the Generation ID and a timestamped frame.

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

  1. 01

    Score your prompt before you generate

    Run your prompt through AVA's pre-flight scoring against the Hand-Anatomy Topology Failure pattern. Green light = generate. Yellow/red = rewrite using the suggested fix before you commit credits.

  2. 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 Hand-Anatomy Topology Failure first appears (e.g. "failure first visible at 1.2s"). Timestamped evidence is significantly stronger than a general complaint.

  3. 03

    Use the correct technical term in your support ticket

    Describe this failure as "Hand-Anatomy Topology Failure". This term maps to a recognised internal workflow in the support system and routes the ticket to the right team.

  4. 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 credits for hand artifacts?

Yes. Submit the Generation ID, identify the failure as a "Hand-Anatomy Topology Failure," and provide a timestamp. Luma's billing support has refunded documented hand failures consistently since the Ray-2 release.

Are hand failures more common on Ray-2 or Dream Machine 1.6?

Ray-2 is significantly better but not solved. The improvement is most visible on static hands and worst on rapid manipulation prompts. Both model versions still fail at hand close-ups regularly.

How do I avoid hand failures on Luma?

Frame hands further from camera. Avoid prompts where hand motion is the primary action. Use short clips (≤4s) when hands are visible. AVA flags hand-risk prompts before submission.

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.

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

Related failures across models

If you’re seeing this failure, you may also encounter these on other models:

Pick a different tool for Luma failures

Some prompt shapes will keep failing on Luma. Routing those shots to a different vendor is the cheapest fix.