Plimo

Hailuo AI (MiniMax): Make Hailuo 3.0 Videos with Sound from Text, Photos or Clips

Pet Dance

What is Hailuo AI?

Hailuo AI is the MiniMax AI video line from the Shanghai company MiniMax, and recent generations have pushed hard on people and physics: Hailuo 02 on difficult moves like gymnastics, Hailuo 2.3 on body movement and facial micro-expressions. Hailuo 3.0 is the name the Hailuo app uses for MiniMax H3, and Plimo offers two versions, MiniMax H3 and MiniMax H3 Max, the faster tier built on it. Both always add sound, run up to 15 seconds, and can work from a prompt alone, from start and end frames, or from reference photos and reference clips.

  • MiniMax H3 Max

    Realistic people and faces.

    • 5โ€“15 s
    • 480P / 768P / 1080P
    • Always with sound
    • Up to 9 reference photos
    • Up to 3 reference clips

    From 105 credits

  • MiniMax H3

    Up to 2K with sound, and can follow reference videos.

    • 4โ€“15 s
    • 768P / 2K
    • Always with sound
    • Up to 9 reference photos
    • Up to 3 reference clips

    From 120 credits

From Hailuo 01 to Hailuo 3.0

  1. September 2024

    Hailuo 01

    MiniMax's first video model, also known as video-01, launched with the Hailuo website and made 6-second clips from a prompt.

  2. June 2025

    Hailuo 02

    Rebuilt on a new architecture MiniMax calls Noise-aware Compute Redistribution, with sharper native output, 6- or 10-second clips and much stronger physics in scenes like gymnastics.

  3. October 2025

    Hailuo 2.3 and 2.3 Fast

    Smoother body movement, more convincing micro-expressions and more styles, from anime to ink wash. The Fast version cut the cost of batch work.

  4. July 2026

    MiniMax H3, or Hailuo 3.0

    One model for text, images, video and audio together: up to 15 seconds with stereo sound where voice, effects and music are generated jointly, multi-shot planning, legible text and brand names, and motion transfer from a reference clip. MiniMax then released its weights.

  5. September 2026

    MiniMax H3 Max

    A tier built on H3 and tuned for text-to-video and photo animation.

How to make a Hailuo AI video on Plimo

  1. Choose H3 Max or H3

    The generator above opens on MiniMax H3 Max, the faster tier built on H3. Switch to MiniMax H3 in the model menu for its 2K quality setting and, once you add reference photos, a Fit photo shape.

  2. Describe the scene, sounds included

    Both versions need a prompt. Write the action, the expressions and what should be heard, like an old fisherman laughing, wiping his eyes and waving at a boat leaving the harbor while gulls call.

  3. Add frames or references

    Use a start frame with an optional end frame, or switch to references and add up to 9 photos plus up to 3 reference clips that together last no more than 15 seconds.

  4. Pick length and shape, then press Generate

    Set the length, choose a shape from ultra-wide 21:9 to vertical 9:16, and sign in. Generate shows the credit cost on the button before you start.

  5. Download or remake

    The video shows up in My creations. Save it, or use Remake to bring the prompt, uploads and settings back for another version.

Why use Hailuo on Plimo

  • Both Hailuo versions under one account

    Send the same prompt to MiniMax H3 and to H3 Max, compare them side by side in My creations, and keep whichever handles your scene better.

  • The rest of the video lineup is a click away

    If a scene needs more than 15 seconds, switch to Kling or Wan in the same model menu without starting over.

  • Pay for the seconds you use

    Hailuo is priced by length and quality, so a short test costs far less than a full take, and the button tells you the exact amount before anything runs.

  • Clean, commercial videos for members

    Hailuo clips created while you're a member download without a watermark and are free to use for a brand, a shop or client work.

  • Borrow motion and faces in one go

    Up to 3 reference clips and 9 photos fit into a single generation, so you can take a move from footage and a face from a photo for the same video.

Hailuo next to other video models

Hailuo's edge is lifelike people and following reference clips. Other families trade that for length, price or a different way in.

  • Kling

    Kling 4.0 doubles the length to 30 seconds and can edit an uploaded video directly, though it takes a single reference clip where Hailuo takes three.

  • Veo

    Veo 3.1 Fast has a polished, cinematic look with sound, but every clip is 8 seconds and only wide or vertical.

  • Seedance

    Seedance 2.5 takes the same mix of photos and up to three reference clips but stretches to 30 seconds and can be muted, while Hailuo always plays sound.

  • Grok Imagine

    Grok Imagine 1.5 has no start or end frame and no reference clips, building its 6- to 15-second videos from up to seven reference photos instead.

  • Wan

    Wan 3.0 accepts even more references, up to ten photos and five reference clips, runs to 30 seconds and can be muted; the Wan family also has an image model.

  • Vidu

    Vidu Q4 Preview can animate a photo without a prompt and handles up to fifteen reference photos, but it doesn't take reference clips.

  • FLUX

    FLUX 3 Video's draft mode gives quick, cheap previews before a final clip of up to 20 seconds, and the family makes still images too.

  • SkyReels

    SkyReels V4 sticks to text, frames and up to three reference photos, with Std and Fast versions that share one set of options.

  • HappyHorse

    HappyHorse 1.1 focuses on combining up to nine reference photos into one video and has no end frame or reference clips.

What to make with Hailuo

  • Faces that carry the scene

    Ask for a close-up of a toddler tasting a lemon for the first time: the scrunched nose, the shiver, then the brave second bite.

  • Copy a move from a clip

    Add a short clip of a dance or a skateboard trick as a reference clip, plus a photo of the character who should perform it, and describe the swap. Motion transfer is one of H3's headline features.

  • Ultra-wide cinema shots

    Choose 21:9 for a letterboxed look, such as a rider crossing a desert ridge at sunset with hoofbeats and wind on the soundtrack.

  • Signs and labels that read right

    MiniMax says H3 renders text and brand names accurately, so try a neon sign spelling your shop's name as it flickers on at night.

  • A short scene with several shots

    Describe a cut from a wide shot of a train platform to a close-up of two friends hugging. H3 was trained to plan multiple shots inside one clip.

  • A pet with attitude

    Hailuo AI image to video works well for pets: start from a photo of your cat and ask for an unimpressed stare, a slow blink and a huge yawn, with a soft purr underneath.

  • A product reveal with sound

    Turn a handful of product photos into a 10-second reveal with the click of a lid and a soft whoosh, ready for a shop page when you make it as a member.

Questions about Hailuo

No. Plimo is an independent website that runs MiniMax's Hailuo models through an API. We aren't affiliated with MiniMax, and credits or plans from the Hailuo app don't carry over to Plimo.

Yes. MiniMax calls the model H3, and the Hailuo app calls it Hailuo 3.0. MiniMax H3 Max is a separate tier built on H3 and tuned for speed.

Hailuo needs a membership. As a new account you get 100 free credits once, enough for a 5-second Seedance 2.0 Mini or 6-second Grok Imagine 1.5 video, and the plans are on the pricing page.

Each version card above shows a starting price. The final number on the Generate button depends on length and quality, and reference clips or extra reference photos add to it. Failed runs are refunded.

MiniMax H3 makes 4 to 15 seconds, and MiniMax H3 Max makes 5 to 15 seconds.

MiniMax H3 Max offers 480p, 768p and 1080p, and MiniMax H3 offers 768p and 2K. Higher quality costs more credits per second.

Always. Both versions add sound to every clip and there's no mute option, so describe what should be heard in the prompt.

Yes. Next to up to 9 reference photos you can add up to 3 reference clips, no more than 15 seconds combined, as MP4 or MOV files up to 50 MB. An end frame only works together with a start frame.

Six shapes from ultra-wide 21:9 to vertical 9:16, plus Fit photo on MiniMax H3 when you use reference photos. With a start frame, the video follows your photo's shape.

Hailuo videos generated during a membership have no watermark and can be used commercially. Videos made without an active membership get a small watermark and are for personal use.

More AI video tools

All models