This is the fast reference for writing MiniMax Hailuo 2.3 prompts — the physics-leading 2026 video model. Keep it open in a tab, copy the pieces you need, and assemble a shot with the formula below. Every table maps to one part of a good prompt, so you can build a clean clip in under a minute.

For full sets, see our best Hailuo prompts roundup and the ready-made templates.

Advertisement

The Hailuo prompt formula

Write one concise line in this order — roughly 40 to 60 words — and Hailuo gives you a controllable clip. The bracket camera moves always come first; everything else follows in plain language.

[camera move(s)] + shot size + subject + action + setting + lighting/style.

[Push in] Medium shot of a barista pulling an espresso shot in a sunlit cafe, steam rising over polished brass, warm morning window light, shallow depth of field, cinematic.

Match each element to a row in the tables below. Skip an element and Hailuo fills the gap for you — usually with something generic — so specify what matters to your shot.

Bracket camera moves

Director camera moves go in square brackets at the very start of the prompt. You can combine up to three, and Hailuo applies them in the order you write them. Two moves are usually cleaner than three.

Bracket tagWhat it doesExample phrase
[Push in]Camera drives toward the subject[Push in] close-up of a face
[Pull out]Camera retreats to reveal context[Pull out] wide reveal of a valley
[Pan left]Rotate view leftward on a fixed point[Pan left] across a market stall
[Pan right]Rotate view rightward on a fixed point[Pan right] along a shelf of books
[Truck left]Slide the whole camera left[Truck left] past a row of windows
[Truck right]Slide the whole camera right[Truck right] along a runner
[Pedestal up]Raise the camera straight up[Pedestal up] from floor to face
[Pedestal down]Lower the camera straight down[Pedestal down] to a tabletop
[Tilt up]Angle the lens upward[Tilt up] a skyscraper facade
[Tilt down]Angle the lens downward[Tilt down] to muddy boots
[Zoom in]Optical tighten on the subject[Zoom in] on a watch dial
[Zoom out]Optical widen from the subject[Zoom out] from a lone figure
[Tracking shot]Follow a moving subject[Tracking shot] of a runner
[Static shot]Locked-off, no movement[Static shot] of a still portrait
[Shake]Handheld-style motion energy[Shake] in a chase sequence

Shot sizes

Shot size sets how much of the subject fills the frame. Pick one and put it right after the bracket move.

Shot sizeFramesUse it for
extreme wide (ELS)Tiny subject in vast spaceLandscapes, scale, establishing
wide (WS)Full subject plus environmentContext and action
medium (MS)Waist upDialogue, product-in-hand
medium close (MCU)Chest upEmotion with some setting
close-up (CU)Face fills the frameExpression, reaction
extreme close-up (ECU)Eyes, mouth, one detailTension, texture, product macro
Advertisement

Lighting looks

Lighting sets the mood. Add one cue near the end of the prompt for a cinematic feel.

LookPhrase
Warm sunsetgolden hour backlight
Cool twilightblue hour ambient light
Bright and cleanhigh-key soft daylight
Dark and moodylow-key single-source light
Studio productsoftbox lighting with rim
In-scene glowpractical neon signs and lamps
Beams and hazevolumetric light, god rays
Edge separationbacklight rim on the subject

Settings quick-reference

Choose these in the app or API, not in the prompt text. The one rule to remember: 1080p is capped at 6 seconds, so any 10-second clip renders at 768p.

SettingOptions & rules
Resolution768p or 1080p. 1080p is capped at 6s.
Duration6s or 10s. Use 768p for any 10s clip.
Aspect (Text-to-Video)Defaults 16:9; also 9:16 and 1:1.
Aspect (Image-to-Video)Follows the source image you upload.
VariantStandard, or Fast (image-to-video focused, quicker and cheaper).
Quality tierPro for top quality, Standard for drafts.
AudioNone — Hailuo outputs silent video. Add sound in an editor.

Style & lens vocabulary

These terms shape the optical look. Combine one lens cue with one style cue at the end of the prompt.

TermPhrase
Wide lens24mm wide-angle
Natural lens35mm natural perspective
Portrait lens85mm portrait compression
Background blurshallow depth of field
Wide flaresanamorphic lens flare
Analog texturefilm grain, 35mm film look
Slow-motionslow motion, high frame rate feel
Documentaryhandheld documentary style

What to avoid

These are the common mistakes that break a Hailuo shot, with the fix for each.

AvoidFix
More than three bracket movesCap it at three; use two for the cleanest result
Bracket tags buried mid-promptPut all brackets at the very start
Two actions in one clipKeep to one main action per shot
Asking for 10s at 1080pDrop to 768p for 10s, or shorten to 6s for 1080p
Writing aspect ratio into a T2V prompt for I2VIn Image-to-Video the source image sets the aspect
Expecting spoken dialogue in the fileAdd voiceover and SFX in an editor afterward
Vague subject like "a person"Name age, clothing, and one distinctive detail
Overloading with adjectivesStay around 40–60 words; trim filler

Example prompts

Six complete prompts built with the formula. Copy, swap the subject, and set the output options in the app.

1. Cinematic b-roll

[Push in] Wide shot of morning fog drifting through a pine forest, thin sunbeams cutting between the trunks, dew on the ferns, golden hour backlight, volumetric light, 24mm wide-angle, cinematic.

A slow push into a wide establishes the space. (16:9, 1080p, 6s)

2. Product hero

[Zoom in] Extreme close-up of a matte black wireless earbud case on wet slate, water beading on the lid, slow rotation of light across the surface, softbox lighting with rim, shallow depth of field, premium.

Best rendered as image-to-video from a product still to lock the shape. (1:1, 1080p, 6s)

3. Action tracking shot

[Tracking shot][Shake] Wide shot of a trail runner sprinting down a rocky coastal path, dust kicking up behind, ocean crashing below, bright high-key afternoon sun, handheld documentary style, energetic.

Tracking plus shake keeps the subject centered with kinetic energy. (16:9, 768p, 10s)

4. Vertical social clip

[Pedestal up] Medium shot of a young chef plating a dessert in a modern kitchen, powdered sugar drifting down, warm practical lamps overhead, shallow depth of field, 35mm natural perspective, appetizing.

A rising pedestal reveal reads well in a 9:16 feed. (9:16, 768p, 10s)

5. Image-to-video motion

[Static shot] Portrait of a woman by a rain-streaked window, subtle motion as she turns her head and blinks, rain running down the glass behind her, low-key single-source light, backlight rim, 85mm portrait compression.

Upload the still first; the aspect ratio follows your source image. (aspect from image, 1080p, 6s)

6. Dialogue setup

[Push in] Medium close shot of an older detective leaning back in a dim office chair, cigarette smoke curling upward, low-key light from a single desk lamp, film grain, tense and weary mood.

The clip is silent — add voiceover and room tone in an editor. (16:9, 1080p, 6s)

Once these feel natural, level up with our realistic-video guide.

Frequently Asked Questions

How many camera moves can I combine at once in Hailuo?

Up to three bracket moves in one prompt. Hailuo applies them in the order you write them, so [Push in][Pan left][Tilt up] runs push, then pan, then tilt. Two is usually cleaner than three.

Do the bracket camera moves go first in a Hailuo prompt?

Yes. Director bracket tags like [Push in] or [Tracking shot] go at the very start of the prompt, before the shot size and subject. Hailuo reads them as camera instructions.

Should I use 768p or 1080p in Hailuo 2.3?

Use 1080p when you need the sharpest output, but note it is capped at 6 seconds. Use 768p for faster, cheaper drafts and for anything that has to run 10 seconds.

6s or 10s — which duration should I pick?

Pick 6s for a single tight action or a product beauty shot, and 10s when you need a fuller motion arc or a reveal. Remember 1080p only supports 6s, so 10s clips render at 768p.

Does Hailuo 2.3 generate audio?

No. Hailuo outputs silent video with no native audio. Add music, sound effects, or voiceover in a video editor after the clip is generated.

How does aspect ratio work in Text-to-Video vs Image-to-Video?

In Text-to-Video you choose the aspect ratio, defaulting to 16:9 with 9:16 and 1:1 also available. In Image-to-Video the aspect ratio follows the source image you upload.

Advertisement