Generate cinematic videos from text or images with precise motion, consistent subjects, and production-ready output.
Text-to-Video creates from a prompt. Image-to-Video animates a source image as the starting frame.
Describe the scene, subject, motion, camera direction, mood, and details to preserve.
Token cost updates automatically based on duration and resolution.
10-second videos cost twice as many tokens as 5-second videos.
Sign in to spend tokens
Prompt is required
Your generated video will appear here
Luma Ray 3.2 generates cinematic video from text or images with precise motion, consistent subjects, and up to 1080p output in 12 aspect ratios. Prizmad ships it as a built-in tool starting at 5 tokens per clip, on the same token balance as every other video and image model.
Luma Ray 3.2 is Luma's video generation model, positioned around cinematic control: precise camera motion, consistent subject details, and production-ready output. On Prizmad it runs in two modes — Text-to-Video builds a clip from a prompt alone, and Image-to-Video animates an uploaded starting frame (JPG, PNG, WebP, or AVIF up to 10 MB). Clips are 5 or 10 seconds long at 540p, 720p, or full 1080p.
Two features stand out for ad production. First, the aspect-ratio list is unusually broad — twelve options from standard 16:9, 9:16, and 1:1 through 21:9 cinematic to banner-friendly extremes like 3:1 and 1:3 — so you can generate for a placement's native shape instead of cropping. Second, up to 4 optional reference images can guide subject consistency or visual style, keeping a product or a look stable across generations. There's also a seamless-loop switch for 5-second clips that blends the end back into the start, made for backgrounds and endlessly looping feed assets.
On Prizmad, Luma Ray 3.2 lives in the same AI Studio as Seedance 2.0, , Happy Horse, image tools, , voiceover, and music — , no Luma account or API key to manage. Generate a clip, and it lands in your asset library ready for the video wizard, captions, or direct download.
Low-angle reveals with slow dolly-ins, precise light sweeps, and crisp reflections — with reference images keeping the packaging consistent with the real product.
Handheld-feel scenes with natural subject motion and warm daylight in native 9:16, built for TikTok, Reels, and Shorts without a creator shoot.
Seamless 5-second loops in wide ratios like 21:9, 2:1, or 3:1 give landing pages and site headers ambient motion that never visibly restarts.
The extreme ratios — 3:1, 1:3, 2:1, 1:2 — let you generate display and banner-shaped video directly instead of awkwardly cropping a 16:9 master.
Reuse the same 4 reference images across a batch of generations to hold one product and one visual style steady through an entire campaign's worth of variants.
Built for precise camera direction — dolly-ins, orbits, light sweeps — with controlled subject movement and consistent details like packaging, faces, and outfits across the clip.
Choose 540p for fast iteration, 720p HD for review cuts, or 1080p Full HD for final production assets. Token cost updates automatically with resolution and duration.
From 16:9, 9:16, and 1:1 through 21:9 cinematic to 2:1, 3:1, 1:2, and 1:3 — generate natively for feeds, stories, banners, and site headers instead of cropping one master.
Attach up to 4 optional subject, style, or product references to keep a real product accurate or hold a visual style consistent across a batch of generations.
The loop switch blends the end of a 5-second clip back into its start — ideal for hero backgrounds, product-page ambience, and endlessly looping feed posts.
Generate from a prompt alone, or upload a JPG, PNG, WebP, or AVIF starting frame up to 10 MB and let the model animate it with described motion.
Open the Luma Ray 3.2 tool in Prizmad's AI Studio and pick a mode: Text-to-Video for prompt-only generation, or Image-to-Video to animate an uploaded starting frame.
Write your prompt — describe the scene, subject, motion, camera direction, mood, and the details you want preserved (packaging text, face, outfit).
Choose the aspect ratio from the 12 options, matching the placement you're producing for rather than cropping later.
Set resolution (540p, 720p, or 1080p) and duration (5 or 10 seconds). The token cost updates as you adjust — 10-second clips cost twice as much as 5-second ones.
Optionally open advanced settings to enable the seamless loop (5-second clips only) or attach up to 4 reference images for subject or style consistency.
Click Generate. Clips typically return in about 2–6 minutes and land in your asset library.
Luma Ray 3.2 is listed at 5 tokens for the default configuration, and the cost scales with your settings: higher resolution raises it, and a 10-second clip costs exactly twice as much as a 5-second clip. The workspace shows the exact cost before you generate.
The model is unlocked on the same token balance as Seedance 2.0, Bernini-R, image tools, avatars, voiceover, and music — one token balance instead of a separate Luma account.
If your tokens run out mid-campaign, a one-off top-up refills the balance directly. Every generated clip is yours with full commercial rights.
On Prizmad it's included in every account rather than offered free-standing. Generations draw from your token balance, starting at 5 tokens for the default configuration and scaling with resolution and duration.
The catalog lists 5 tokens for the default setup. Cost scales with resolution (540p, 720p, 1080p) and duration — a 10-second clip costs twice as many tokens as a 5-second clip at the same resolution. The workspace recalculates the exact price as you change settings.
Clips are 5 or 10 seconds at 540p, 720p, or 1080p, in any of 12 aspect ratios from 9:16 vertical through 21:9 cinematic to banner shapes like 3:1 and 1:3.
It blends the end of the clip back into the start so playback repeats without a visible cut — useful for hero backgrounds and looping feed assets. It's available for 5-second videos only.
Yes. You can attach up to 4 optional reference images for subject, style, or product guidance. Reusing the same references across a batch helps hold packaging, colors, and look steady through multiple clips.
Luma Ray 3.2 reaches 1080p, offers 12 aspect ratios including banner shapes, and has a seamless-loop option, with clips up to 10 seconds. Seedance 2.0 tops out at 720p but goes to 15 seconds, supports up to 9 tagged reference images, and can generate audio. Both run on the same token balance, so testing them head-to-head costs nothing extra in setup.
Yes. Videos generated on Prizmad carry full commercial rights — use them in paid ads, on landing pages, and across social and owned channels with no extra licensing.
The workspace estimates about 2–6 minutes per clip, varying with resolution and duration. Generations run asynchronously, so you can queue several variants and compare results.
Published 2026-07-16 · Last updated 2026-07-16