Luma connects creative planning, image editing and generated video

Luma AI is a creative platform built around image and video generation, editing and agents that help organise the work. Its current Luma app goes beyond the earlier Dream Machine generation interface: teams can keep references, concepts and iterations together on a visual board. The useful question is whether that shared workspace improves the path from an approved brief to usable media.

01 / ProductA creative workspace as well as a model provider

Luma’s new app guide² describes boards that contain references, media, versions and project context. Agents can route work across Luma and third-party models, while creators direct the result through instructions and controls. The platform therefore combines an organising interface with access to generation capabilities.

The current Agents API FAQ⁵ identifies Uni-1 and Uni-1-max for images and Ray3.2 for video. These are the models exposed through that API, which is a distinct product surface from every model available inside the creative app. A buyer should specify whether they want a collaborative workspace or a programmatic generation service.

Older Dream Machine tutorials and subscription references remain online. They are useful when working in that product, but should not be treated as the current price list for every Luma service. This blueprint covers one company and distinguishes its current app and Agents API where the workflow or billing differs.

02 / AudienceFor creative teams that can judge and refine the result

Luma is a candidate for concept development, campaign imagery, short video exploration and variations of existing creative material. It is most useful when the team has a clear brief and can recognise what should remain consistent: the product, character, visual language, message or sequence of events.

The platform does not remove creative direction. Someone must decide which concept serves the audience, which details are accurate and what makes a shot usable in the final edit. A model can produce an attractive image that fails the brief by changing packaging, inventing text or showing an action that the product cannot perform.

Start with work where iteration is practical and a reviewer can inspect the output. A short visual concept is easier to evaluate than a large campaign generated from an unsettled brief. For exact typography, repeated layout production or a tightly controlled filmed performance, compare conventional design and production tools as part of the same decision.

03 / WorkflowApprove the concept before spending heavily on motion

A proposed pilot could be a short product introduction with several supporting still images. Add the approved product photography, copy, audience and delivery formats to a board. Define which elements must be preserved and which can change. Keep the intended final use visible so that an appealing experiment does not become the project by accident.

Use the Research Mode guide³ when the direction needs development. It describes guided questions, reference gathering and planning documents such as briefs and shot lists. Research Mode itself does not generate media; production moves into Create Mode. That creates a useful opportunity to agree on the concept before consuming generation credits on polished variations.

Generate and review still frames first. Check product shape, text, composition and the relationship between shots. If the source is a finished flat image, the Layers guide⁴ explains how to extract elements into an editable stack with transparent backgrounds. This can make a local revision more manageable than regenerating the whole composition. Review the extracted boundaries and text before treating the result as a finished editable design.

Then animate the approved direction using the appropriate video operation. The Agents API supports generation, editing and reframing as distinct request types. Keep the source frame or video, prompt and settings with each candidate. When a variation fails, identify whether the problem is motion, identity, composition or the brief itself before trying another prompt.

Finish by reviewing the actual export in its destination format. Inspect the whole clip, not only the first frame. Check transitions, product continuity, small text and any audio relationship. Record the accepted assets and the revision history so the team can repeat the useful parts of the process. This is a suggested production test, not a report of private Luma use or measured campaign results.

04 / PricingSeparate app credits from API generation prices

Luma’s app pricing¹, checked on 15 September 2026, lists Plus, Pro and Ultra with 10,000, 40,000 and 150,000 credits respectively, alongside business plans. Its detailed usage table varies by model, resolution and operation. The Agents API pricing guide⁶ instead quotes US dollars per generation, with separate provisioned-capacity arrangements.

The API pricing guide says video rates are subject to change ahead of general availability. For an illustrative set of twenty five-second 720p SDR API candidates, generation totals US$6 at the listed rate. If only two are accepted, that is US$3 per accepted clip before other work. This simple calculation is more useful than treating every attempt as finished footage. Include image preparation, edits, reframing and final production in the real budget.

The app’s Layers guide and current pricing table describe costs differently for higher-resolution extraction. Use the current pricing table and the confirmation shown in the app for the operation being purchased, rather than assuming an older tutorial’s multiplier. Record model and resolution alongside usage so the team can explain a sudden increase in credit consumption.

05 / DistinctionsA shared board can make iteration easier to understand

The main workflow benefit is the ability to keep the reference, instruction and candidate near one another. That can make creative review more concrete: a reviewer can point to the intended product image and the variation that changed it, rather than reconstructing context across unrelated downloads and chat messages.

Local editing also changes the economics of revision. If a background is acceptable but the headline needs work, changing the selected element can preserve more of the approved composition. The benefit depends on the quality of the extraction and edit, so evaluate that with the kinds of assets the team actually produces. A complex image with overlapping objects is a different test from a simple poster.

Our Runway blueprint is a relevant comparison for generated video and creative production controls. Use the same brief, reference assets and acceptance criteria to compare the resulting workflow. Our HeyGen blueprint is more relevant when the central requirement is a presenter or localised spoken video. Choosing the right production approach can matter more than comparing isolated image quality.

Conventional filming, stock footage and design software remain useful alternatives or complements. A generated establishing shot may fit one brief, while a precise demonstration needs real footage. Evaluate Luma as part of the production process that delivers the required result, including the finishing tools and human review that remain necessary.

06 / QuestionsConsistency and export handling are the practical limits to test

A good single frame does not prove continuity across a sequence. Review repeated products, characters and environments together. Look for changes in shape, labelling, lighting and spatial relationships. Define which differences are acceptable creative variation and which make an asset unusable before generating a large batch.

For an API integration, the delivery lifecycle matters too. The FAQ describes asynchronous jobs and temporary output URLs. The application should wait for a completed state, download the accepted output to controlled storage and retain the generation identifier. A temporary provider URL should not become the permanent asset link in a customer-facing product.

The same documentation notes that repeated image prompts are not deterministic through a public seed parameter. Preserve accepted source assets and use references when consistency is important; do not assume that a saved prompt alone will recreate an identical image later. This is particularly relevant when a client requests a small change to an earlier approved campaign.

Finally, separate visual approval from factual and usage approval. Check logos, claims, people and source material against the brief and the team’s permissions. Keep the review proportionate to the destination: an internal mood board and a published product advertisement have different acceptance standards. The production process should make that distinction explicit.

07 / DecisionChoose Luma for an iterative creative process with clear review

Luma is worth evaluating when a team wants to keep creative context, generation and revision in one workspace, or needs an API for media generation. Pilot a complete brief with a fixed review process and track the cost of accepted output. That shows whether the platform improves the work the team actually delivers.

Compare specialist presenter tools, conventional production or other video platforms when their controls better match the brief. Prepare references and acceptance criteria when the creative direction is still unsettled. The strongest use of Luma combines rapid exploration with disciplined selection and finishing.