NOLGIA

Guides / Models

MiniMax H3 on NOLGIA: Hailuo 3, H3 Max and what each one takes

NOLGIA carries three MiniMax video lines. Hailuo 3 is the flagship: 2K, 4 to 15 seconds, six aspect ratios, up to nine reference images and three audio tracks, or a start and an end frame, with stereo sound and speech rendered in the same pass. H3 Max is the budget line in two variants, one for text with references and one for a start frame with an optional end frame. Hailuo 2.3 is the older, silent, 16:9 model. This guide runs one input on each and shows what came back.

  • 9 min read
Hailuo 3FramesText

What you make, in order. Tap one to jump to its step.

H3 is the model behind most of the talking presets on NOLGIA, from the social ad to the outfit try-on, and the one Cover reveal uses to land exactly on your still. It earns that by taking more kinds of input in one run than most models: a person from a photo, a voice from an audio clip, a start and an end frame, and a spoken line, with the sound made alongside the picture.

The subject here is a premium ad frame, a woman in a white suit in a white studio under one orange light, made with an image model. Each render below uses a different input on a different variant, so read the captions for which is which.

The family, and what each one takes

ModelQualityLengthInputsSound
Hailuo 3768p or 2K4 to 15 sText; or up to 9 reference images and 3 audio tracks; or a start and an end frame. Frames and references do not mixStereo, with speech
H3 Max, text variant480p or 768p5 to 15 sText, up to 9 reference images, 3 audio tracks. No framesYes, no speech
H3 Max, image variant480p or 768p5 to 15 sA start frame (required), an end frame (optional). No referencesYes, no speech
Hailuo 2.3768p or 1080p6 or 10 sText or a start frame, 16:9 onlySilent
Hailuo 2.3 Fast768p or 1080p6 or 10 sA start frame (required), 16:9 onlySilent
From the public model catalog on 2026-09-24. Hailuo 3 and H3 Max render 21:9, 16:9, 4:3, 1:1, 3:4 and 9:16. All on Pro and up. The Hailuo 3 and H3 Max pages are the live versions.

In the video composer the family appears as Hailuo 3, H3 Max, Hailuo 2.3 and Hailuo 2.3 Fast. H3 Max switches between its two variants on its own: add a start frame and it runs the image variant, leave it out and it runs the text variant.

What goes wrong

MistakeWhy it failsFix
A start frame and a reference image together on Hailuo 3The provider takes frames or references, not both.Frames for an exact opening and ending, references for a new shot.
A spoken line on H3 MaxH3 Max renders sound but not speech.Move to Hailuo 3 for a line.
A 4 second clip on H3 MaxIts range starts at 5 seconds.Ask for 5, or use Hailuo 3 for 4.
A vertical clip on Hailuo 2.3It renders 16:9 only.Use Hailuo 3 or H3 Max at 9:16.
Writing the sound as an afterthoughtWithout a Sound: clause the model invents the track.One Sound: line that names the room and the loudest thing in it.
Expecting 2K on H3 MaxH3 Max tops out at 768p.Hailuo 3 is the 2K tier.

Before you start

  • Decide the input first: a photo of a person as a reference, a start and an end frame, or words alone. The variant follows from it.
  • For a spoken line, use Hailuo 3 and write the line in quotes.
  • For a start and end frame, make the two stills first. The end frame should be the same place and light a few seconds later.
  • Hailuo 3 defaults to 2K; pick 768p for a draft. The Generate button shows the exact price.

Hailuo 3: a subject reference and a spoken line

Pick Hailuo 3, attach the still under Reference images and address it as @Image1. Write what she does, the line she says in quotes, and the sound. We rendered at 2K, 16:9, 5 seconds.

A woman with slicked-back dark hair in a tailored white double-breasted suit standing in a white studio, a hard orange light from the side throwing a long shadow across the floor
The reference still. Made with an image model for this guide.
The prompt, as sent
The woman in @Image1, in her white suit, in the same white studio lit by one hard orange light. She takes three slow steps towards the lens, stops, and says, low and certain: "Now you're looking." The wind machine lifts the ends of her hair, the long shadow sweeps across the floor as she moves. Medium shot, slow push in, no cuts, no new people. Sound: the hum of a wind machine, her steps on the studio floor, the line clean and close, no music. No text, no logos.
A woman in a white suit walks from the back of a white studio towards the camera under a hard orange side light, stops in a medium shot and speaks
Hailuo 3, 2K, 16:9, 5 s, one reference image, one run. The studio, the suit and the orange floor light are read from @Image1. She walks in from the back wall to a medium shot and the line lands in the last second: the track goes loud in the speech band right there. We measured the band rather than transcribing the words.
The video composer on a desktop with Hailuo 3 selected, the still attached under Reference images as @Image1, the composer's hint on addressing attachments, 2K chosen and the price on the Generate button
Reference images on Hailuo 3. The still rides as @Image1 and the composer tells you so under the slots.
The same Hailuo 3 run on a phone with the reference image attached
On a phone. The reference slot and the hint sit under the prompt.

Try it yourself

  • The reference still (JPEG, 16:9)JPEG photo · 1536x864 · 58 KBDownload
  • The end frame (JPEG, 16:9)JPEG photo · 1376x768 · 36 KBDownload
Hailuo 3, Reference images, 2K, 16:9, 5 seconds, the prompt above. The end frame is for the next step.

H3 Max: a start frame and an end frame

Pick H3 Max, choose Animate an image and add the still as the start frame. Open the end frame slot under it and add the second still. The composer says what the slot does: it pins the clip's final frame while the start frame stays the start. The clip has to travel from one to the other, so the prompt only needs to say how.

Start frame

Start frame: the woman in the white suit facing the camera in the white studio

End frame

End frame: the same woman a few steps to the right, side-on to the camera, looking back over her shoulder, the hem lifting
Two stills, one place. The end frame was made from the start frame with an image model, so the studio and the light match.
The prompt, as sent
She walks slowly from where she stands to the right of the frame and turns side-on, glancing back over her shoulder at the lens as she goes, the hem of the jacket lifting in the wind, her long shadow sweeping across the white floor with her. Camera locked off, no cuts, no new people. Sound: the hum of a wind machine and her steps on the studio floor, no music. No text, no logos.
The woman in the white suit turns from facing the camera, walks to the right and ends side-on, glancing back, the orange shadow sweeping across the floor
H3 Max, 768p, 16:9, 5 s, start and end frame, one run. It opens on the first still, she walks right and turns side-on, and the last frame settles on the second still's pose, glancing back over her shoulder. The shadow travels with her. Sound is the wind machine and her steps, with no speech, which is what H3 Max makes.
The video composer on a desktop with H3 Max selected, the white studio still as the start frame and the second still in the end frame slot, the price on the Generate button
The two slots. Start frame above, end frame below, one prompt, the price on the button.
The same two slots on a phone, the end frame row saying it pins the final frame
On a phone. The end frame row sits under the start frame.

H3 Max: words alone

No attachments, so H3 Max runs its text variant. We described the same studio in words at 21:9, 768p, 5 seconds, to see what the budget line builds from nothing.

The prompt, as sent
Premium advert, photoreal, 21:9. A seamless white cyclorama studio. A woman in a tailored white double-breasted trouser suit walks slowly from the left edge of frame to the centre, one hard orange light from the right throwing her long shadow across the white floor ahead of her, a wind machine lifting the hem of the jacket. She stops in the centre and turns her head to the lens. Wide shot, camera locked off, no cuts. Sound: the hum of a wind machine and her steps on the studio floor, no music. An invented person. No text, no logos.
A widescreen clip of a woman in a white suit walking from the left edge of a white studio to the centre and turning to the camera, an orange light throwing her shadow across the floor
H3 Max text variant, 768p, 21:9, 5 s, one run. No attachments. She walks from the left edge to the centre and turns to the lens; the white studio, the orange floor light and the long shadow all came from the words. The light is softer than the still's hard gel, which is the difference a reference makes.

How to write for H3

  • Name each attachment for its job. The woman in @Image1, the voice in @Audio1. The composer labels the slots for you.
  • Say the shot in film words. A slow orbit, a push in, a locked off wide. A named move gives a steadier camera path than a mood.
  • The line in quotes, on Hailuo 3 only. H3 Max makes sound but does not speak.
  • One Sound: clause. It shapes the whole track: the room, the loudest thing, and no music if you mean it.
  • Frames or references, not both. Decide before you attach.

Where H3 already works for you

You do not have to open the composer to use it. These presets run on Hailuo 3 and fill the prompt from a form: Cover reveal pins your still as the final frame, You on camera films a saved character with a real camera move, Social media ad, Creator review ad and App ad make talking clips, and Outfit try-on reads a fit head to toe.

What it costs

Per clip, by tier and by second, sound included. Hailuo 3 defaults to 2K; H3 Max defaults to 768p. The Generate button shows the exact charge for the settings on screen.

  • Hailuo 3Pro and up

    Per 5s clip

    768p 37 credits · 2K 65 credits

  • H3 MaxPro and up

    Per 5s clip

    480p 14 credits · 768p 23 credits

  • Hailuo 2.3Pro and up

    Per 5s clip

    768p 28 credits · 1080p 41 credits

  • Hailuo 2.3 FastPro and up

    Per 5s clip

    768p 16 credits · 1080p 28 credits

Read live from the catalog when this page loads. Every price is shown before you generate. See every rate.

Which one for which job

JobModelWhy
A person from a photo says a lineHailuo 3, referencesSpeech and stereo sound in the same pass
A clip that must open and close on two exact stillsH3 Max image variant, or Hailuo 3Both take a start and an end frame; H3 Max is cheaper, Hailuo 3 is 2K
A silent product turn from one photoH3 Max image variantA start frame at the lowest price in the family
Words alone, many referencesH3 Max text variantUp to nine images and three audio tracks without frames
A 16:9 clip with no sound at allHailuo 2.3The only silent one

Checklist

  1. Input decided: references, frames, or text.
  2. Variant follows: Hailuo 3 for a line, H3 Max for frames on a budget.
  3. Length inside the range: 4 to 15 on Hailuo 3, 5 to 15 on H3 Max, 6 or 10 on Hailuo 2.3.
  4. Attachments named in the prompt: @Image1, @Audio1.
  5. A Sound: clause written.
  6. The price read on the button before Generate.

Questions and answers

Is MiniMax H3 the same as Hailuo 3?
Yes. Hailuo 3 is the model's name; H3 is the short form. On NOLGIA the flagship is listed as Hailuo 3 and the budget line as H3 Max.
Which of them can speak?
Hailuo 3. H3 Max renders sound effects and ambience but not a spoken line, and Hailuo 2.3 is silent.
Can I use references and a start frame together?
Not in one run. Hailuo 3 takes frames or references; the composer moves a frame into the references with a notice if you add both. H3 Max splits the two into its image and text variants.
How many references does Hailuo 3 take?
Up to nine images and three audio tracks of up to 15 seconds each. It takes no reference videos.
What resolution do I get?
Hailuo 3 renders 768p or 2K, with 2K as the default. H3 Max renders 480p or 768p. Hailuo 2.3 renders 768p or 1080p.
Which plan do I need?
Pro and up for the whole family. Topping up credits on a lower plan does not unlock them.

Give Hailuo 3 a photo and a line.

Open Hailuo 3

Related guides

Video

Turn a photo into a video, three ways

Animate one still on NOLGIA three ways: the video composer for control, the agent for a hands-off run, a preset for the fastest clip. Which models take a photo, and what each way costs.

· 13 min read