MiniMax H3 for clips that come with sound

MiniMax H3 is a video generation model from MiniMax, the company behind the Hailuo series, and the newer generation of that line. On Swayclip it runs text-to-video, image-to-video and reference-to-video.

Describe the subject, one clear action, camera movement, lighting and the sound you want. With a product photo, describe motion rather than reinventing the product. For references, state what each image contributes. Keep complex hands, hinges and exact label text out of a first test; inspect the output because prompt instructions cannot guarantee fidelity.

It depends on length and resolution. The table on this page lists credits per second and the cost of a full clip, read from live pricing, and the generator shows the exact amount before you run.

Yes. Every clip includes an audio track generated by the model, and it cannot be turned off. If you need a silent clip, pick another model.

In our product-video test it kept shapes steady but showed small deformation on fine parts, such as a charging-case lid. Clips are 6 or 10 seconds long and took a few minutes to render. Treat it as one test, not a lab study.

Paid-plan and Credit Pack outputs can be used commercially within the paid entitlement scope, but you are responsible for rights in uploaded images, logos, and likenesses.

No. Swayclip is an independent creative tool that shows the real underlying model name.

The current standard signup offer is 50 credits for eligible new accounts; eligibility and abuse-prevention rules apply. That is below the cost of the shortest H3 clip shown in the live table, so the standard grant alone does not cover an H3 generation. Review your balance and the previewed cost before submitting. A $9 one-time pack adds 300 non-expiring credits without a subscription; subscriptions are another option. Failed generations refund credits automatically. Free-entitlement outputs are watermarked and for personal exploration.

No. H3 Max is a separate model variant in MiniMax’s official documentation. Swayclip currently exposes MiniMax H3 and its image-reference mode, not H3 Max. Do not assume the speed or resolution claims of H3 Max apply to these results.

No. Swayclip’s H3 reference mode generates a new clip from 2–3 images. It has no source-video upload or verified Recast workflow. Do not use it as a promise of character replacement or preserving an existing performance.

Ref2VA is MiniMax’s official reference-to-audio-video model variant. The official model supports broader multimodal references, including video and audio. On Swayclip, H3 reference mode currently takes only 2–3 images; those broader upload capabilities are not exposed here.

You can ask for box clicks, paper rustle or room ambience in a text prompt; H3 generates audio alongside video. These are generated sounds, not recordings or guaranteed timing. Our product examples test simple photo animation, not a complete unboxing or shop-ambience workflow. Listen to the result before using it.