Minimax H3 Text to Video: pick the row, not the brand (5s, 6s, 7s, 2K, 26cr)

Minimax H3 Text to Video is one catalog generate. Use it because this job matches this row.

MiniMax, text-to-video, 26 credits, max output 2K. Description: "MiniMax H3 is a frontier video model — generates 2K video from a text prompt alone, in durations from 5 to 15 seconds across six aspect ratios." requires_image is false. audio is unset. Durations: 5s, 6s, 7s, 8s, 9s, 10s, 11s, 12s, 13s, 14s, 15s. Aspects: 21:9, 16:9, 4:3, 1:1, 3:4, 9:16. Qualities: 768P, 2K. That is the whole contract. "MiniMax" as a brand also files Hailuo video under a different key. Picking the brand is how you land on the wrong SKU. Picking this row is how you get a 2K clip from words, 5–15s, at 26 credits a second.

Text alone, 2K, 5 through 15 seconds

No still required. No driving video. The prompt is the scene. MiniMax's register is narrative: scene progression, character actions, visual atmosphere — not a hardware-store of lens names. Write one continuous beat sized to the duration you actually selected. 5s is not a trailer. 15s is not a short film. 7s (the title example) is a single action that can resolve.

2K is the listed max. 768P is the other quality. If the brief is 720p social with a locked product still, this is the wrong row — you wanted image-to-video. If the brief is a widescreen scene from a sentence, 21:9 is on this list and many i2v rows will not give it to you from text.

AI video generator is the door. Generate a video from text is the agent. Best text-to-video model is the ranked shelf. There is no MiniMax provider hub on Versely; Hailuo's roster is a different family. This page is H3 text-to-video only.

Pick the row because the job is a scene from words

Use H3 when:

• you do not have a still that must be frame one
• you need a duration in 5–15s, including the awkward in-between lengths (6s, 7s, 11s) that 5-or-10 enums do not offer
• 2K is the plate you actually want
• six aspects, including 21:9, matter

Do not use H3 because a tweet said MiniMax. Do not use it to animate a logo you already have — requires_image is false, and the model will redraw the logo. Do not use it as lipsync. audio is unset; native audio is a different catalog flag.

26 credits is per second on the matrix. A 5-second take and a 15-second take are not the same bill. Pick the length first, then write the beat to fit. Padding a 5-second idea to 15 seconds is how the back half invents a second scene.

Brand is not a workflow

Hailuo sits next to this company in conversation and not on this record. Flux, Kling, Wan, Veo are other rows. H3's distinguishing facts are text-only, 2K, 5–15s, six aspects, 26 credits. If your job is a 10-second still-lock on a product, image-to-video wins. If your job is 7 seconds of a scene that does not exist as a frame yet, this row wins. That is the whole decision.

The text-to-video glossary is the mechanic. This SKU is that mechanic at 2K. Use it when the slate is empty and the length lives in this enum. Use something else when the slate already has a still.

FAQ

Does Minimax H3 Text to Video require an image?

No. The description is explicit: 2K video from a text prompt alone. If you have a still that must survive as frame one, you are on an image-to-video row, not this one.

What lengths can I actually set?

Every second from 5s to 15s. The title names 5s, 6s, 7s because those are the short end of a range most 5/10 enums cannot hit. 26 credits meters per second, so the length is a cost choice as well as a taste choice.

Is there native audio?

The record does not set audio: true. Do not promise a mixed dialogue plate from this SKU. Picture at 2K. Finish sound elsewhere if the cut needs it.

Why not just pick "MiniMax" in a vendor app?

Because this catalog row is H3 text-to-video at 2K / 5–15s / 26 credits. Other MiniMax-family products are other rows. The job matches this spec or it does not. The brand name will not add an image input or a soundtrack the record does not list.