Hailuo 2.3 Text to Video: Turn Written Scenes into Clips

Last verified: October 19, 2026

Hailuo 2.3 turns a written description into a short video clip. You type the scene, choose how long it runs, and the model renders it without a reference image. That makes Hailuo 2.3 Text to Video a job of description: the model builds the video from your text alone, so the subject, action, setting, light and camera all have to be in the words you write.

MiniMax built Hailuo 2.3 to make motion and performance more convincing. Physical action, stylized looks and small changes of expression are rendered more convincingly, and the model follows instructions about movement more closely. For prompt writing, this means movement is worth spelling out: how a body moves, how a face changes, how an object turns.

Vidofy keeps the form to three controls: the Prompt box, a Video Duration choice, and an Enhance Prompt switch that decides whether the model expands your text before it generates. Below you'll find the kinds of clips the model handles best, how to direct the camera from inside the prompt, and what to change when a run misses.

Explore more text to video

Capability Snapshot

Hailuo 2.3 at a Glance

What goes in, what you can set and what comes out of each run.

Input

A text prompt of up to 2,048 characters

Video Duration

6 or 10 seconds

Output

1 video per run

Prompting Techniques for Hailuo 2.3

Direct the Camera with Bracketed Commands

Write camera moves straight into the Prompt box as bracketed commands, each placed right after the part of the description it applies to. Hailuo 2.3 accepts fifteen: Truck, Pan, Pedestal, Tilt and Zoom in both directions, Push in and Pull out, plus Shake, Tracking shot and Static shot. Commands that share one bracket, such as [Pan left,Pedestal up], happen at the same time. Commands placed one after another through the prompt play in sequence, so "the climber reaches the ledge [Tracking shot], then looks out over the valley [Pull out]" gives you a follow shot and then a reveal.

Switch Off Enhance Prompt When Your Wording Is Final

Leave Enhance Prompt on while you're testing an idea written in a sentence or two, and turn it off once the prompt says exactly what you want. The switch lets the model's prompt optimizer expand your text before it generates, and MiniMax advises turning that optimizer off for more precise control. That matters most when your prompt contains bracketed camera commands, a named art style or a fixed order of actions, because an expanded version could water them down.

How to Generate a Video with Hailuo 2.3

Four steps take you from a written scene to a downloaded clip.

Text-to-Video or Image-to-Video with Hailuo 2.3

When to start from words, and when an existing picture is the better starting point.

Starting point of the shot In text-to-video the model builds the whole shot from your description, with no image involved. Hailuo 2.3 image-to-video takes a picture as the clip's first frame, together with a text prompt. Choose text-to-video when you're inventing a scene from scratch, and image-to-video when a specific picture, such as an approved product photo or a character design, has to open the clip.
Directing camera moves You can write camera moves into the prompt using MiniMax's bracketed commands, which are documented for this model in text-to-video. Image-to-video with Hailuo 2.3 accepts the same fifteen bracketed camera commands in its prompt. Camera control doesn't separate the two, so choose by starting point; commands you've tested in text-to-video use the same syntax if you later move a shot to an image-led workflow.
Use text-to-video when the scene exists only in your head, and Hailuo 2.3 image-to-video when a specific image has to be the opening frame. Camera direction works the same way in both.

Fixing Common Hailuo 2.3 Problems

Common reasons a run misses, and what to change before the next one.

My prompt was rejected for sensitive content

MiniMax screens prompt text for sensitive content, and its error reference gives one fix for flagged input: change the input content. Don't resubmit the same wording. Rephrase or remove the element most likely to have been flagged, keep the rest of the scene as it was, and generate again.

The camera movement looks chaotic or skips some of the moves I asked for

Putting many moves in one bracket asks the camera to do them all at once, and MiniMax recommends no more than three combined commands. Cut the bracket down to one or two moves, or turn them into a sequence by placing each command after the beat it belongs to. Choose the longer Video Duration so the sequence has time to play out, then run it again with only that change so you can see whether it fixed the shot.

My prompt is too long for the Prompt box

The Prompt box has a character limit, and MiniMax documents text prompts for this model at up to 2,000 characters, slightly less than the box accepts. Keep prompts to 2,000 characters or fewer. Cut repeated adjectives and mood words first, keep the subject, action, setting, light and camera commands, then generate again.

I need a video longer than a single clip

Each run renders a clip only as long as the Video Duration you picked, so longer pieces have to be built from several shots. Choose the longer duration, write each shot as its own prompt, generate them one at a time and edit the downloads together. Repeat the same character and style description word for word in every prompt to give the shots the best chance of matching.

Frequently Asked Questions

How much does it cost to generate a Hailuo 2.3 video?

The credit cost of each run is calculated from the options you choose and shown above the Generate button once you're signed in. Video Duration is the setting that changes it, so check the cost after you pick a length and before you click Generate.

Can Hailuo 2.3 show realistic facial expressions and emotion?

Yes. MiniMax made live-action facial performances and micro-expression changes more natural in Hailuo 2.3, and uses small changes of expression to build character performances. To get that from a prompt, describe the expression changing over time, such as a smile fading and the eyes looking away, and frame the face close enough to see it. A fixed label like "sad woman" gives the model nothing to act out.

Does lighting stay realistic when the camera moves?

Hailuo 2.3 keeps lighting direction, shadow changes and color tones close to photorealistic even when the camera moves dynamically. Use that by naming the light source and its direction, such as a low sun from the left or a neon sign behind the subject, whenever you plan a tracking or pushing camera. That gives the model a fixed light to keep consistent as the view changes.

Can Hailuo 2.3 create visual effects shots?

MiniMax lists strong VFX among Hailuo 2.3's strengths, aiming for cinematic realism and immersive effects, alongside style shifts and surreal looks. In the prompt, describe what the effect physically does, such as where it starts, how it spreads and what it lights up, instead of just naming it. That gives the model motion to render.

What frame rate are Hailuo 2.3 videos?

MiniMax lists Hailuo 2.3 at 24 fps, the standard film frame rate. Clips drop into a 24 fps edit without frame-rate conversion.

References

Sources and citations used to support the content provided above.

Updated: 2026-10-19 22:29:29 6 Sources

platform.minimax.io

Source Link
https://platform.minimax.io/docs/guides/video-generation

www.minimax.io

Source Link
https://www.minimax.io/news/minimax-hailuo-23

www.minimax.io

Source Link
https://www.minimax.io/news/minimax-and-veed-hailuo-23-for-pro-level-ai-video

platform.minimax.io

Source Link
https://platform.minimax.io/docs/api-reference/video-generation-t2v

platform.minimax.io

Source Link
https://platform.minimax.io/docs/api-reference/errorcode

platform.minimax.io

Source Link
https://platform.minimax.io/docs/guides/models-intro