Guide6 min readSeptember 13, 2026

How to Use MiniMax H3 Image to Video: Durations, 768P vs 2K and Exact Credit Costs

A practical guide to MiniMax H3 image-to-video on VdoBloom: 4 to 15 second clips with native audio, the real per-second credit rates at 768P and 2K, prompting rules and cheaper alternatives.

MiniMax H3 Image-to-Video on VdoBloom animates a still photo into a 4 to 15 second clip with native audio, priced at 15 credits per second at 768P and 24 credits per second at 2K — so a 6-second 768P clip costs 90 credits ($4.50) and the same clip at 2K costs 144 credits ($7.20). It is the image-to-video half of the H3 family (MiniMax’s Hailuo 03 generation), it carries the highest rating in VdoBloom’s video picker at 4.9, and it is the model to reach for when a clip needs to sound finished as well as look finished.

What it is best at

Three things separate H3 I2V from the rest of the image-to-video lineup:

  • Native audio. The clip comes back with sound generated alongside the picture, not bolted on afterwards. Ambience, room tone and movement noise arrive matched to what is happening on screen.
  • Long single takes. Every whole-second duration from 4 to 15 is available. Most image-to-video models on the platform cap out at 5 or 10 seconds, which forces you to stitch. H3 can carry a 15-second beat in one generation.
  • Cinematic motion from one frame. It handles camera movement and subject movement together rather than choosing one, which is what makes a still photo read as a shot instead of a parallax card.

The trade is cost. H3 is one of the more expensive per-second models on VdoBloom, and at 15 seconds of 2K it is the single most expensive image-to-video render you can launch. Use it where the output is the deliverable, not where you are still exploring.

Real specs and exact credit costs

Durations are 4 to 15 seconds in one-second steps. Resolution is 768P or 2K. The image-to-video variant has no aspect-ratio control — framing is inherited from the photo you upload, which is why your crop matters before you ever write a prompt. The prompt field accepts up to 7,000 characters. Dollar figures use the Lite rate of $15 for 300 credits, so one credit is $0.05.

Duration768P credits768P cost2K credits2K cost
4s60$3.0096$4.80
5s75$3.75120$6.00
6s90$4.50144$7.20
8s120$6.00192$9.60
10s150$7.50240$12.00
12s180$9.00288$14.40
15s225$11.25360$18.00

The pattern is linear: 15 credits per second at 768P, 24 per second at 2K, with no discount for longer clips and no odd jump at any duration. The intermediate seconds not listed above (7, 9, 11, 13, 14) follow the same per-second rate exactly.

Because the rate is flat, duration is your main cost lever. Trimming a planned 10-second clip to 6 seconds saves 60 credits at 768P — three dollars — and most social cuts are better at 6 seconds anyway. See the pricing page for the plans these credits come from; annual Lite is $10 a month, and one-time packs start at $2.49 for 75 credits that never expire.

How to prompt H3 image-to-video well

The photo does most of the work, so the prompt should describe change, not content. Things the model already sees in your frame do not need restating.

  1. Name one motion and one camera move. The subject turns toward the window as the camera pushes in slowly. Two instructions beat six.
  2. Write the sound you want. Because audio is native, saying quiet room tone with distant traffic, no music genuinely changes what comes back.
  3. Lock identity explicitly. Add preserve the facial features, clothing and background exactly as in the uploaded image. Drift is most visible past 8 seconds.
  4. Pace the shot to the duration. A 15-second generation needs a beginning, a middle and an end written into the prompt, or the model loops a gesture and the clip feels padded.
  5. Test at 768P. Prove the motion at 60 credits for 4 seconds before committing to a 2K render of the same idea.

A worked example

Goal: a 6-second vertical clip from a portrait photo, for the top of a reel. Crop the photo to 9:16 before uploading, because the I2V path takes its framing from the source. Then prompt:

The person from the uploaded image lifts their gaze to camera and exhales, shoulders settling, while the camera pushes in slowly from a medium shot to a close-up. Preserve the facial features, hair, clothing and background exactly as in the original image. Late-afternoon light, faint dust in the beam. Audio: quiet interior room tone, one soft breath, no music.

Run it at 6 seconds, 768P: 90 credits, $4.50. If it lands, re-run at 2K for 144 credits and you have a 6-second clip with usable sound for 234 credits total, or $11.70 — under one Lite month. Start from the image to video tool and pick MiniMax H3 in the model list. If your photo shows a real person, you must have that person’s consent before animating it.

When to pick a different model

ModelExample costPick it instead when
Hailuo 2.3 Standard I2V15 credits for 6s at 768PYou are drafting, testing framing, or making volume content where 90 credits a clip is not defensible
Kling 2.5 Turbo40 credits for 5sYou want fluid, natural human motion and do not need generated audio
MiniMax H3 Text-to-VideoSame per-second rateYou have no source photo — and it adds aspect-ratio control, including 21:9
MiniMax H3 Image-to-Video90 credits for 6s at 768PThe clip is a deliverable, needs native sound, or needs to run longer than 10 seconds in one take

A practical workflow that respects the price: block the shot on Hailuo 2.3 Standard at 15 credits, confirm the photo animates the way you expect, then re-shoot the winning idea on H3. You spend 105 credits instead of gambling 90 on a first attempt that may be framed wrong.

All paid plans download watermark-free with commercial rights. Free accounts get 10 credits on signup with no card, which is not enough for an H3 render — use those on a cheaper model first, or see the free AI video generator page for what the free tier realistically covers.

Frequently asked questions

How long can a MiniMax H3 image-to-video clip be?

Four to fifteen seconds, in one-second steps, in a single generation. There is no need to stitch two clips for a 15-second beat.

Does MiniMax H3 image-to-video include sound?

Yes. Audio is generated natively with the video, so you can direct it in the prompt rather than adding a track afterwards.

Can I choose the aspect ratio?

Not on the image-to-video variant — it follows the uploaded photo, so crop before you upload. The text-to-video variant does expose aspect ratios, from 21:9 through 9:16.

What does a 10-second 2K clip cost?

240 credits, which is $12.00 at the Lite rate of $0.05 per credit. The same 10 seconds at 768P is 150 credits, or $7.50.

Is MiniMax H3 the same as Hailuo 03?

Yes — H3 is the Hailuo 03 generation of MiniMax’s video models, listed on VdoBloom under the H3 name with separate text-to-video and image-to-video entries.

Ready to try it?

Create your first AI video in minutes — no credit card required.

Start Creating Free →