← All models

Ray 3.2

Video Available Full comparison ↗

luma/ray-3-2 · by Luma AI · Diffusion transformer (DiT)

Pricing — 1 offering(s)

Luma AI

View on Luma AI ↗

About this offering: Luma AI Ray 3.2, the current API-available Ray model. Ray 2 has been superseded and is no longer separately priced for API access. Video-to-video and extend modes are priced differently (see lumalabs.ai/api). HDR output adds 2× cost.

T2V 5s / 720p

  • $0.30 / clip Current 2026-06-13 → present

T2V 10s / 720p

  • $0.90 / clip Current 2026-06-13 → present

T2V 5s / 1080p

  • $1.20 / clip Current 2026-06-13 → present

T2V 10s / 1080p

  • $3.60 / clip Current 2026-06-13 → present

Showing the active price and any recorded history. Full pricing history is available via the paid API — see API docs.

See how Ray 3.2 fits into a cost-aware routing setup

See how →

Capability profile

How Ray 3.2 rates across core capability dimensions, with the task-level evidence behind each rating.

camera control Strong

Luma AI's signature strength; camera movements respond reliably to cinematic terminology (dolly, pan, orbit) in the text prompt.

temporal consistency Strong
motion quality Strong
Prompt adherence Moderate
native audio Weak

No audio generation.

clip duration ceiling Moderate

Max 10 seconds; 5s and 10s billed separately.

image to video quality Strong

Strong I2V mode that preserves reference composition.

Resolution ceiling Moderate

Max 1080p; no 4K via API.

Inference speed Moderate

Cloud API; per-clip billing with generation times typically under 60s for 5s clips at 720p. 1080p and longer clips take longer. Competitive with peers; no dedicated "fast" tier.

cost efficiency Moderate
video editing Moderate

API exposes both V2V (style/reference-guided regeneration of existing footage) and extend (temporal extension of an existing clip via prompt). The extend mode in particular constitutes directed modification of an existing video beyond I2V start-frame anchoring. No inpainting or mask-guided element replacement.

Ratings are estimated — limited independent data is available for this model.

Operator guidance

First choice when predictable camera motion is essential. Per-clip pricing makes cost easy to forecast ($0.30 for 5s 720p, $0.90 for 10s 720p). 1080p is significantly more expensive ($1.20/$3.60 for 5s/10s) — use 720p unless resolution is critical. No audio; choose Veo 3 if audio is needed.

Use cases

Limitations

Citations