Luma AI pricing (formerly Dream Machine): plans & Ray credits

Product and naming references: Luma's official product reference

Magic Hour publishes this pricing guide and offers alternative image and video workflows. Luma product, plan, credit, and API claims are based on the linked official Luma sources checked September 13, 2026; this is not a sponsored review or controlled output-quality benchmark.

Current Luma App plans and free access

The official Luma pricing page was checked September 12, 2026. Its paid cards begin with Plus; Luma's subscription guide separately describes limited free access. Confirm the allowance, model access, watermark, commercial rights, local taxes, and billing period in your account before generating.

Luma pricing page, captured September 25, 2026. Billing selection, currency and any promotion are shown in the screenshot; recheck the live page and checkout before buying. View official source

Plus includes access to Luma and supported third-party image and video models, guest collaboration and commercial use. Higher tiers increase usage capacity. The current pricing page lists both Team and Enterprise business offers. Luma App and Luma API have separate billing; a web subscription is not an API allowance. Open the browser workspace: Luma AppLuma API

Luma API pricing: Ray3.2 costs in dollars

The official Luma API page lists the following base SDR rates, checked October 2, 2026. These are API prices per request, separate from the browser app’s credit table.

HDR is priced at twice SDR; HDR plus EXR at three times SDR. The Build offer has no minimum commitment, but rate limits apply and it has no latency SLA. The page does not specify a standing free API allocation; confirm any trial credit in the developer account rather than transferring a free web allowance.

Start an integration with Luma’s current video API documentation. Its current developer surface differs from historical Dream Machine examples. Record the exact model, operation, requested duration, resolution and output format next to every quote.

For example, ten billed five-second 720p SDR generation requests at $0.30 total $3.00 before other operations. If five pass your brief, generation-only cost is $0.60 per accepted clip. This is illustrative arithmetic, not an observed acceptance rate. Count actual billed attempts and finishing work before comparing providers.

Why older Dream Machine pricing tables differ

Dream Machine is a legacy Luma name that still appears in older URLs, help articles and search queries. Luma's current official reference says to use Luma App for the browser workspace, Ray3.2 for the current video model, UNI-1.1 for the current image model, and Luma API for developer access. Historical pages may list Free, Lite, Plus or Unlimited plans with different allowances. Identify the date, product, model and billing route before comparing quotes.

How much does one Luma video cost?

The current credit table varies by model, operation, resolution and output format. For example, Ray3.2 text-to-video or image-to-video in base SDR is listed at 100 credits for five seconds at 720p, or 400 credits for five seconds at 1080p. HDR and EXR modes have separate multipliers; video-to-video has separate rates.

Those are credits per attempt. They do not include the cost of retrying an unsatisfactory result or making additional edits. Always check the current estimate for the selected settings.

A defined production budget

Suppose you need twenty five-second Ray3.2 clips at 720p SDR and budget three attempts for each accepted clip. At 100 credits per attempt, that is 6,000 credits. The current 10,000-credit Plus allowance covers that illustrative generation budget, leaving capacity for other work.

At the $30 monthly Plus price, allocating the entire 10,000-credit balance evenly gives $18 of the subscription value to those 6,000 credits. Your invoice is still $30, and the calculation assumes those settings and three attempts per accepted clip. It is not a measured success rate or a promise that twenty usable clips will result.

If you change to the listed 1080p setting, the same sixty attempts cost 24,000 credits. That is why resolution and retries matter more than a generic “cost per video” estimate.

Which plan should you choose?

Choose the smallest plan that covers your measured workload and required features. Plus can be sufficient for a small project; regular publishing alone does not establish that you need Pro. Consider Pro or Ultra when your actual model/settings budget exceeds the lower allowance. Consider Team for shared billing and administrative controls, not just because several people review your clips.

Before committing annually, generate a short sample with representative inputs and record the credits spent on every attempt.

When does using another tool save money?

Only when the complete workflow costs less for an acceptable result. Drafting in one platform and finishing in another can add export steps, extra generations and inconsistent visual styles. It is an option to evaluate, not a proven universal saving.

For a new shot, compare the same source and output brief in Magic Hour Image-to-Video or Text-to-Video. If your job is to transform footage you already have, use our video-to-video guide to compare Luma Modify with other editing workflows. Include retries, finishing time, and the actual export in the cost. Use the AI video pricing worksheet to record retries and accepted clips; the AI video generator comparison covers the broader platform choice.

Price one finished Luma sequence

A finished sequence usually costs more than one generation. List shots, candidates per shot, generated duration, resolution, extensions, modifications and upscales. Use the live Ray mode and plan quote for every line item; do not reuse a credit number from an older Dream Machine table.

For a four-shot sequence with three candidates per shot, budget twelve initial generations. Add extensions only to accepted candidates. If continuity at a join fails, count the replacement attempt. Divide total credits or dollars by the shots that pass the brief and can enter the edit.

Test the hardest shot first: a recognizable person or product, a camera move, an occlusion, and the delivery aspect ratio. If that shot cannot preserve the required subject, a larger subscription increases attempt volume without fixing workflow fit.

When another workflow is cheaper

Use Luma when its current generation or modification workflow matches the shot. Compare another platform when the same project also needs text-to-video, image-to-video, face replacement, lip sync or audio work and transferring assets creates extra time.

Run one accepted-output comparison in Magic Hour Image-to-Video. Keep source, prompt, duration and acceptance criteria constant. Record all attempts and repair time. Pick the route that delivers the usable shot, not the route with the lowest advertised monthly price.