Cost intent · official pay-as-you-go rates checked August 4, 2026
Budget MiniMax H3 video experiments for an Unreal game, account for references and retries, and spend only after a playable slice proves the idea.
MiniMax’s official pay-as-you-go page lists H3 at $0.13 per output second for 2K and $0.08 per output second for 768p as checked August 4, 2026. A 15-second output is therefore $1.95 at 2K or $1.20 at 768p before chargeable extra images, reference video, regenerations, rejected takes, editing, storage, review, and production implementation. Budget by accepted decision, not generated clip. When the unresolved question is whether the mechanic is fun or controllable, put the next dollar into a SEELE Unreal vertical slice.
Choose the prompt closest to the user job, review the prefilled scope, and generate the native project. Each prompt converts a video-model intent into input, state, objective, failure, restart, and an inspectable SEELE handoff.
Create a native Unreal 5 third-person puzzle slice in one original research chamber. The player aligns three energy nodes, receives clear feedback, unlocks an exit, can enter an invalid state and recover, can fail, and can restart immediately. Include browser preview and downloadable project output for an accepted-decision cost comparison.
Use this Unreal promptGenerate an Unreal 5 exploration scene with one controllable route, one camera transition, one interaction, completion, failure recovery, immediate restart, browser preview, and native project download. Keep the world small so the team can compare learning against further video-generation spend.
Use this Unreal promptCreate a compact Unreal 5 objective flow with a functional tracker, one interaction prompt, one success state, one error state, controller and keyboard navigation, restart, browser preview, and local project output. Use original assets and deterministic runtime text.
Use this Unreal promptBuild an Unreal 5 vertical slice with one signature movement mechanic, three short challenges, one finish, fall recovery, restart, browser preview, and a downloadable project. Keep scope under five minutes and record the tested build as the source for future captures.
Use this Unreal promptAn inspectable Unreal project rather than a rendered clip, prompt answer, or unsupported provider-integration claim.
A fast playable check for camera, controls, objective clarity, feedback, success, failure, and restart before local continuation.
A workflow for performance review and package preparation before external distribution or platform submission.
A local handoff for source, Blueprints, assets, plugins, rights, build settings, target-device checks, and continued production.
The official table lists $0.13 per second for 2K and $0.08 per second for 768p. Four seconds costs $0.52 or $0.32; ten seconds costs $1.30 or $0.80; fifteen seconds costs $1.95 or $1.20 at those dated rates.
Official pricing says audio input is free, the first five images are free and additional images cost $0.04 each, while input video is billed by duration and output resolution. Re-check current rules before production.
The cited page lists 768p-to-2K regeneration at $0.05 per output second and notes that original input materials are billed again. A cheap first pass can still become expensive when selection criteria are vague.
H3 is economical when a clip resolves a visual or motion decision. It is waste when repeated clips avoid the harder question: can a player understand, control, finish, fail, and restart the promised experience?
Name the one decision, the maximum number of attempts, duration, resolution, reference allowance, review owner, acceptance condition, and stop condition. Do this before generating the first clip.
Calculate output seconds, chargeable images, reference-video seconds, expected regenerations, 2K conversion, storage, transfer, editorial time, rights review, and the production work required after approval.
Use the lowest resolution and duration that can answer the question, then promote only the accepted take. Preserve prompts and input versions so a useful direction can be reproduced without blind retries.
If the decision concerns interaction, difficulty, readability, camera control, state, or performance, create a SEELE native Unreal slice instead of more video. Measure the cost of learning, not the cost per clip alone.
Track attempts, accepted output, review time, downstream rework, and whether the artifact changed a production decision. Use accepted-decision cost to route the next task between H3, Unreal implementation, or another method.
Quoting a dated per-second rate as a guaranteed future or regional price.
Ignoring reference and regeneration charges, reviewer time, storage, editing, and downstream implementation.
Optimizing cost per clip while producing no accepted decision or playable evidence.
Using high-resolution long clips before framing, motion, continuity, or mechanic scope is approved.
| Dimension | What good looks like | Evidence to keep |
|---|---|---|
| Accepted-decision cost | Spend is tied to one approved production decision | Attempts, seconds, inputs, review time, and result |
| Stop discipline | The experiment ends when it passes, fails, or reaches its ceiling | Predeclared limit and final disposition |
| Downstream value | The output reduces rework in art, animation, cinematic, UI, or marketing | Owner confirmation and next artifact |
| Prototype alternative | Interactive questions move to a native build promptly | SEELE slice cost and learning record |
Estimated API spend equals output seconds times the dated output rate, plus chargeable reference images, input-video seconds, regeneration, and any other documented charges. Keep labor and infrastructure in separate visible lines.
A $1.95 output is not a $1.95 decision when nine attempts fail. Measure the full accepted set, including review and recreation work, so cheap generation does not hide expensive selection.
Approve composition, motion, duration, continuity, and content at the cheapest adequate setting. Use 2K only when fine detail or delivery truly requires it.
Use H3 for video questions. Use a native Unreal project for input, collision, systems, performance, saves, networking, packaging, or player experience. SEELE shortens that switch by generating the bounded project.
Snapshot date: August 4, 2026. MiniMax’s July 31, 2026 release and current API documentation identify the API model as MiniMax-H3 and support H3’s video modes, duration, resolution, frame rate, multimodal references, audio, and dated pricing. They do not establish an Epic-supported Unreal integration or native game output. Epic documentation and the tested project remain authoritative for engine behavior.
First-party release positioning, H3 technologies, multimodal context, video and audio claims, editing, motion transfer, and target industries.
Open official announcementModel name, supported generation modes, 4–15 second duration, resolution, frame rate, aspect ratios, and reference inputs.
Re-check current output and reference-material pricing before spending. Use Epic documentation and the actual project for engine, asset, animation, UI, cinematic, build, and package decisions.
Compare MiniMax H3 and Seedance 2.5 on matched game-video tasks, then move the winning idea into a native Unreal 5 project in SEELE.
Open this workflowConvert a MiniMax H3 game concept into a scoped mechanic, generate a native Unreal 5 project in SEELE, and validate the playable loop.
Open this workflowUse MiniMax H3 for cinematic previs, then rebuild approved shots in Unreal Sequencer with controllable scenes, cameras, timing, and evidence.
Open this workflowAs checked on August 4, 2026, MiniMax’s official pay-as-you-go page lists $0.13 per output second at 2K and $0.08 per output second at 768p. Pricing, regions, taxes, quotas, and terms can change, so verify the official page before approving a budget. Confirm the current provider documentation before you lock a production decision.
At the dated output rates, fifteen seconds is $1.95 at 2K or $1.20 at 768p. That calculation excludes additional reference images, input-video charges, regeneration, repeated attempts, storage, editing, rights review, and the production work needed to recreate an approved idea in Unreal. Treat the tested Unreal build, not the reference clip, as acceptance evidence.
The official pricing page says audio input is free and the first five images are free, with $0.04 for each additional image. Input video is billed by duration and output resolution. Re-check the current table because material billing and model availability can change. Keep prompts, outputs, project versions, and review notes in the same decision log.
Switch when the unresolved question depends on player input, collision, state, navigation, AI, UI data, difficulty, performance, saves, networking, or packaging. Another clip cannot prove those. Use the accepted visual direction in a bounded SEELE prompt and test the native Unreal result. Recheck rights, runtime behavior, performance, and packaging before any external release.
Do not compare a video-second price with a project price in isolation. Compare the cost to answer the decision. H3 may cheaply resolve framing or motion; SEELE may more directly resolve whether the proposed mechanic, objective, failure, restart, and player experience work in a native Unreal project. Use the browser preview to reject weak interaction before deeper production work.
Track cost per accepted production decision: API charges, all attempts, reference charges, regeneration, storage, reviewer time, edits, recreation, and downstream rework. Also record whether the decision reached a tested Unreal artifact. A low clip price with no accepted learning is not efficient. Record the source clip separately from the native project and package evidence.
Use H3 to answer the video question. Use SEELE to generate the playable system: native Unreal project, browser preview, optimization and packaging path, and a downloadable handoff. Keep concept media and native evidence clearly separated.