How much does MiniMax H3 cost for ad creative?
How much does a 6-, 10-, or 15-second MiniMax H3 ad creative cost as of August 1, 2026? Official rates put a 6-second clip at $0.78 and a 15-second 2K spot at $1.95, but closed-beta 768p, missing video-package support, and the 15-second generation cap make list price a ceiling, not what a working pipeline pays.
- Platform
- Multi-platform
- Campaign type
- Short-form paid social
- Spend range
- $7.80-$1,950/month (API fees)
- Timeframe
- 2026-08-01
- API fee per second (2K)
- 0.13
- Verdict
- mixed
- Last reviewed
- 2026-08-01
As of August 1, 2026, MiniMax H3 costs $0.78 in API fees for a successful 6-second 2K generation and $1.95 for a successful 15-second 2K ad spot at the official pay-as-you-go rate of $0.13 per second. That is cheap enough to change a testing calendar. It is not the working price of a complete ad-production pipeline, because it excludes concepting, editing, review, approval, media spend, and any extra successful generations needed to get a usable cut.
Price snapshot: MiniMax H3 pay-as-you-go rates on August 1, 2026
| Billing item | Official MiniMax H3 pay-as-you-go price or rule | Planning consequence for ad creative |
|---|---|---|
| 2K video output | $0.13 per second | Use this as the current launch-planning rate for most H3 ad tests. A 6-second clip is $0.78; a 15-second spot is $1.95 in API fees. [1] |
| 768p video output | $0.09 per second; closed beta | The cheaper tier exists in the docs, but it should not be the default budget number unless the buyer actually has access. [1] |
| Input video | Billed at the same rate as the output video | Video-to-video workflows do not get a separate discount line; model the output tier and duration. [1] |
| Input images | First five input images are free, then $0.04 per image | Most simple product-reference requests will not move the budget much, but image-heavy workflows need a small extra line item. [1] |
| Input audio | Free | Audio input does not add an API fee, though editing, rights, and mix work still sit outside this rate card. [1] |
| Failed or security-review generations | Not billed | The invoice should not charge for those failed/security-review calls, but the production schedule can still lose time. [1] |
| Maximum generation length | 15 seconds per generation | Good for short ads and bumpers; binding for longer cuts unless the team stitches multiple generations. [1] |
For ad creative, the useful H3 pricing number is not a subscription headline. It is the per-second API fee multiplied by the duration of each successful output. That puts H3 in a different planning category from a conventional shoot: the generation call itself becomes small enough that the bottleneck moves to prompts, selection, brand review, post-production, and platform trafficking.
What a 6-, 10-, or 15-second H3 ad costs
| Ad unit | Typical use | 2K API fee at $0.13/sec | 768p API fee at $0.09/sec, if accessible |
|---|---|---|---|
| 6-second bumper | Hook test, product flash, short retargeting asset | $0.78 | $0.54 |
| 10-second creative | One message, one visual idea, one CTA | $1.30 | $0.90 |
| 15-second short ad | Full short-form paid social unit within H3’s cap | $1.95 | $1.35 |
| 9:16 vertical variant | Reels, Shorts, TikTok-style layout | Same duration math; a 10-second 2K vertical is $1.30 | Same duration math; a 10-second 768p vertical is $0.90, if accessible |
Aspect ratio is a creative and platform-spec decision, not a separate public price tier in the cited pay-as-you-go sheet. A 9:16 vertical ad should be budgeted by output tier and seconds unless MiniMax publishes a different billing rule.
The image-input rule is small but worth modeling honestly. If, hypothetically, a team uses eight reference images for one generation, the first five are free and the remaining three would add $0.12 in API fees before the video-generation charge. That is not a budget breaker; it is the kind of line item that gets missed when a rate-card screenshot becomes a media plan. [1]
This is the same cost-per-creative discipline used in our GPT API versus dedicated ad-creative tools pricing framework: isolate the variable generation fee first, then add the labor and workflow costs that decide whether the system is actually cheaper.

Monthly testing math: 10, 100, and 1,000 H3 creatives
For a growth team, the meaningful question is not whether one clip costs less than lunch. It is what happens when the calendar asks for enough variants to test hooks, offers, formats, angles, and landing-page matches. At the official 2K rate, the API-fee line stays small even at volumes that would have been painful under traditional production assumptions. [1]
| Monthly output volume | 6-second bumpers at 2K | 10-second creatives at 2K | 15-second shorts at 2K |
|---|---|---|---|
| 10 successful generations | $7.80 | $13.00 | $19.50 |
| 100 successful generations | $78.00 | $130.00 | $195.00 |
| 1,000 successful generations | $780.00 | $1,300.00 | $1,950.00 |
That table is the economic event. A small agency could put 100 successful 15-second 2K generations into a month for $195 in H3 API fees. At 1,000 successful 15-second generations, the direct generation bill is $1,950 before the rest of the workflow. For teams that already have a creative strategist, editor, and buyer in place, the generation fee is unlikely to be the largest constraint.
The phrase “successful generations” is doing work. If one approved ad requires multiple usable generations, the invoice scales with each billed output. If a reviewer rejects the concept, if the edit needs a human clean-up pass, or if the platform versioning takes extra time, none of that is visible in the $0.13-per-second number.
Third-party production-economics references are useful here only as outside context. Hedra’s AI-generated advertising materials frame generative video as a way to increase ad variation and production throughput, but that is vendor positioning, not evidence that H3 will improve a specific account’s CPA. [2] ngram gives a more concrete directional contrast, citing roughly $4,500 per minute for traditional production versus about $400 per minute for AI video, and $10,000 to $50,000 for a 30-second traditional spot; those figures are not MiniMax first-party pricing and should not be treated as proof of H3 performance. [3]
The narrower, safer conclusion is still enough: H3’s listed API fees make it inexpensive to create a large batch of candidate ad assets. Whether those candidates become profitable ads depends on creative judgment, iteration speed, post-production quality, and media buying.

The three gates that keep list price from becoming pipeline price
The 768p tier is not a default launch budget
The $0.09-per-second 768p line is real in the official pay-as-you-go documentation, but it is marked as closed beta. That matters because a buyer who models 1,000 15-second clips at $1.35 each instead of $1.95 each is assuming access that most launch plans should not assume. [1]
For current planning, the cleaner rule is to budget H3 at 2K pricing unless the account has confirmed 768p access. If 768p opens later, the savings can be treated as upside rather than a surprise shortfall.
MiniMax video packages do not support H3 yet
MiniMax also publishes video-point packages priced from $1,000 to $6,000 per month, but the official video-packages page says those packages do not support H3 yet. A team cannot assume package economics will lower the effective H3 rate until MiniMax adds H3 support to that program. [4]
That does not make the pay-as-you-go price unattractive. It simply removes a common procurement shortcut. If the next month’s plan is built on H3, use the H3 pay-as-you-go rate rather than blending in package discounts that do not currently apply.
The 15-second cap is a format constraint
The 15-second maximum generation length fits a lot of paid-social work: hooks, product demos, short offer tests, vertical variants, and retargeting cuts. It is less convenient for teams that need 30-second, 45-second, or longer story structures. Those teams can still explore H3, but the pipeline now includes stitching, continuity management, edit review, and probably more approvals. [1]
For a 15-second ad, the cap and the billing unit line up cleanly. For anything longer, the $1.95 figure stops being the ad price and becomes the cost of one segment.
Launch claims, reseller rates, and open weights are modifiers—not the current price
MiniMax’s own H3 launch post says H3 is priced at “less than a third of mainstream rivals.” That is useful launch positioning, but it is not a clean benchmark unless the comparison set, public rates, output settings, and billing assumptions are matched. For this pricing record, the official MiniMax H3 pay-as-you-go table is the anchor. [5]
Reuters reported on July 31, 2026, that MiniMax planned to release H3’s weights within days. That could become important for infrastructure economics, self-hosting choices, and vendor dependence, but it is not a current ad-creative cost option in this pay-as-you-go model. Treat it as a future variable and compare it against the broader open-weight ad-creative benchmark context once there is an actual release and operating cost to model. [6]
Atlas Cloud’s reseller-verified review lists H3 at $0.14 per second for 2K and $0.10 per second for 768p. That is slightly above MiniMax’s official $0.13 and $0.09 pay-as-you-go rates, which makes it useful as markup context rather than the base rate for direct MiniMax planning. [7]
There is also a circulating early-tester figure of roughly $1 for a 15-second clip. It should not replace the official math. The public MiniMax pay-as-you-go rate puts a 15-second 2K H3 generation at $1.95 in API fees, and that is the number a buyer can actually tie back to the current official documentation. [1][8]
Operating rule for August 2026 H3 ad planning
For the July 31, 2026 launch record, keep the MiniMax H3 launch Tracker entry separate from the budget model: launch timing explains why teams are looking at H3; the pay-as-you-go sheet explains what the generation call costs today.
- Use $0.13 per second for current 2K H3 planning.
- Treat $0.09 per second 768p as unavailable unless the buyer has confirmed closed-beta access.
- Do not model MiniMax video-package discounts for H3 until MiniMax explicitly adds H3 support.
- Treat $1.95 as the API fee for one successful 15-second 2K generation, not the total cost of a finished ad.
- Model longer ads as multiple segments plus editing and review, because the current generation cap is 15 seconds.
H3’s list-price API cost is extremely low for ad testing. The working pipeline price is still gated by access, packaging, duration limits, and the labor around the generation call.
References
- Pay-As-You-Go — MiniMax.
- AI Generated Advertising — Hedra.
- Who Makes Best AI Generated Ad Videos — ngram.
- Video Packages — MiniMax.
- MiniMax H3 — MiniMax.
- China's MiniMax releases H3 video model — Reuters, July 31, 2026.
- MiniMax H3 Review — Atlas Cloud.
- What Is Hailuo H3 — kie.ai.
Built on this evidence
No Bidding tactic or Creative record currently cites this case file. Compare it against other results in Benchmarks.
Related benchmark reading
Report a corroborating or contradicting result
Seeing something different in your own account? Feed the data-integrity loop instead of leaving an open comment.