MiniMax H3 (Hailuo) AI Video Generator
MiniMax H3 is the video generation model from MiniMax, the company behind Hailuo AI
What is MiniMax H3?
MiniMax H3 is the video generation model from MiniMax, the company behind Hailuo AI. On blubiu it comes in two versions: H3, which renders 768p or 2K and can take reference images and reference videos, and H3 Lite, which is far cheaper per second and can make clips as short as 1 second. Neither version generates sound.
Developer: MiniMax · Checked 2026-11-02
What MiniMax H3 is good at
- +2K output: the full H3 is the only video model on blubiu that renders 2K.
- +Following a reference video: H3 accepts up to 3 reference videos (billed per second of reference footage) alongside up to 9 reference images.
- +Very cheap drafts and very short clips: H3 Lite costs a fraction per second and goes down to 1 second, which suits testing motion before a final render.
- +Shots built around one character: give it a consistent character sheet plus a few pose or scene stills as references, so each clip starts from the same look.
- +Picture-only footage you will score or voice later: the output is silent, so there is no generated music or dialogue to strip out in the edit.
When not to use it
- −You need sound in the clip. H3 and H3 Lite are silent; Seedance 2.0 and Veo 3.1 can generate sound with the picture. See Seedance 2.0 →
- −You want a mid-price 720p or 1080p clip with references. Seedance 2.0 covers that range. See Seedance 2.0 →
- −The person in a photo must perform the exact moves from a real clip. AI motion transfer takes the moves straight from the footage. See Kling Motion Control →
MiniMax H3 versions and specs
| Version | Inputs | Length | Resolution | Sound | Price |
|---|---|---|---|---|---|
| MiniMax H3 Lite (Hailuo) | Text, First frame (+ optional last), References (images ≤9, audio ≤3) | 1–15 s | 480P / 768P | Silent | 480P 2 · 768P 4 credits/s, 1 s minimum; subscribers 40% off: 480P 1.2 · 768P 2.4 credits/s |
| MiniMax H3 (Hailuo) | Text, References (images ≤9, videos ≤3, audio ≤3) | 5–15 s | 768P / 2K | Silent | 768P 19.4 · 2K 29 credits/s, 5 s minimum; subscribers 40% off: 768P 11.64 · 2K 17.4 credits/s; reference video 768P 16.2 · 2K 27 credits/s |
Standard prices, read live from the platform catalog; once signed in, the Generate button shows your account's actual price. Failed generations are refunded.
How to prompt MiniMax H3
Draft on Lite, render on H3
Lite and H3 read prompts the same way, so get the motion right on Lite at 480p first, then switch to H3 for the 768p or 2K version.
Keep reference videos short
Reference footage is charged per second on top of the output, so trim it to the part that shows the motion you want.
Plan the sound separately
Because the output is silent, write the prompt purely for picture and add voice-over or music afterwards (blubiu's voice-over node works for this).
Frames on Lite, references on H3
Only Lite can open on a first frame and end on a last frame, and a run uses either frames or reference images, not both. The full H3 has no frame mode: in the generator, click the tag on your image to switch it from frame to reference.
Say what each reference is
With several references, open the prompt by naming them ("image 1 is the character, image 2 is the street, video 1 is the camera move"), then describe the action. The model then knows which file each instruction is about.
Frame for 16:9 or 9:16 on Lite
On blubiu, Lite renders landscape 16:9 or vertical 9:16: portrait ratios such as 3:4 are sent as 9:16, and everything else, square included, as 16:9. Crop a first frame to that shape before you upload it so nothing important is cut.
MiniMax H3 FAQ
Is MiniMax H3 the same as Hailuo AI?
Hailuo AI is MiniMax's own app; MiniMax H3 is a video model from the same company. On blubiu you call the model directly and pay with blubiu credits — you are not using the Hailuo app or its account.
Does MiniMax H3 generate sound?
No. Both H3 and H3 Lite produce silent video. If the clip needs ambient sound or dialogue, use Seedance 2.0 or Veo 3.1, or add audio afterwards.
How is it priced?
Per second of output at the chosen resolution; H3 adds a per-second charge for any reference video you pass in. Current rates are in the table above, and the Generate button shows the total before you confirm. Failed generations are refunded.
H3 or H3 Lite?
Use Lite to explore motion and timing cheaply, or when a clip shorter than 5 seconds is enough. Use H3 when you need 768p/2K quality or a reference video.
Can I try it for free?
Yes, with the Free plan's monthly credits. Lite at 480p is the cheapest run on this page, the Generate button shows the exact cost before you confirm, and failed runs are refunded automatically. Plans and credit packs are on the pricing page.
Can H3 Lite follow a reference video?
No. Reference videos work only on the full H3; Lite takes text, a first and last frame, or reference images. Draft the motion on Lite from text or stills, then add the reference video when you switch to H3 for the final render.
Can it start from my own image?
Yes, in two ways. Lite takes your image as the first frame, optionally with a last frame, which is image to video. The full H3 takes your images as references instead, up to 9, so the clip follows them but doesn't necessarily open on that exact picture.
How long can a clip be?
Both versions go up to 15 seconds per run, and Lite goes down to 1 second; each version's exact range is in the table above. For a longer piece, generate several clips with the same references and join them in your editor.
Can I get 4K?
Not directly: 2K is the highest it renders. To go further, upscale the finished clip with FlashVSR on the AI video upscaler, which keeps its motion, length and frame rate. The same works for a Lite draft you want to keep at 1080P.
MiniMax H3 or Seedance 2.0?
Pick H3 for 2K or for a silent clip guided by a reference video. Pick Seedance 2.0 when the clip needs sound, when you want a first and last frame at 1080p, or when a mid-price 720p clip is enough.
Tools that use MiniMax H3
Create Video with AI →
Create video with AI from a sentence, a first frame, first and last frames or up to 9 reference images: 1–15 second clips with Seedance, Veo, MiniMax and Wan.
AI Text to Video Generator →
Turn a written description into a 1–15 second video with Seedance, MiniMax H3, Veo and more — no images needed.
AI Image to Video Generator →
Upload an image as the first frame, describe the motion and generate a video with Seedance, Veo 3.1, MiniMax H3 Lite or Wan 3.0.
Pair MiniMax H3 with
Consistent Character Sheet Generator →
Make the reference image H3 needs: one sheet with three face angles and a full turnaround of your character.
Image Generator AI →
Draw the first frame for Lite, or extra stills of the set and props to pass to H3 as references.
AI Video Upscaler →
Take a 2K render to 4K, or a Lite draft you like to 1080P, without generating it again.
AI Motion Transfer →
When a real person must copy the exact moves from a dance clip, motion transfer takes them straight from the footage.
Other video models
Make 4–15 second videos with sound from text, a first frame or up to 9 reference images with ByteDance's Seedance 2.0 — Mini, Fast and 2.0 compared.
Generate an 8-second 720p clip with optional sound using Google's Veo 3.1 Fast, from text or from a first and last frame. One fixed price per clip.
Use Alibaba's Wan 3.0 for 2–15 second videos with sound from text, a first frame or reference images, or to replace the person in a video.
Run Kling Motion Control 2.6 online: a full-body photo copies the moves from a 3–30 second video, at 720P or 1080P. No Kling account needed; try it free.
Upscale an MP4 or MOV up to 120 seconds to 720P, 1080P, 2K or 4K with FlashVSR, a video super-resolution model. Billed per source second.
See image models →