Hailuo T2V-01 Director
MiniMax's T2V-01-Director reads camera moves written in square brackets, like [Push in] or [Pan left], and follows them in a 720p clip of about 6 seconds.
Overview
T2V-01-Director is the camera-control model in MiniMax's 01 series. Next to the scene description, its prompt can carry camera instructions in square brackets, and the model makes each move at the point in the text where the instruction sits. Output is a landscape clip at 720p.
The default prompt shows how this works. It opens with [Push in] on a close-up of a tense woman in a darkened kitchen who is startled by a sound, then adds [Shake] as her shock turns to fear. In the sample, the camera closes in on her face during the first second or so. The clip stays a single continuous black-and-white shot, with hard shadows and pans hanging just above her head.
Choosing a variant: Hailuo 01 Text to Video is the same series without camera commands, for prompts where the model decides the framing. Hailuo 02 Text to Video is the later generation. MiniMax lists it as accepting these same bracketed commands, and it also offers 1080p and 10-second options. Pick Director when you want planned camera moves at 720p from the 01 series, for example so a new shot cuts cleanly into an edit built from earlier T2V-01 clips.
What you can build
- Previs and storyboard apps where the user picks a camera move from a menu and the app writes the matching bracket into the prompt.
- Trailer and teaser shots with a planned reveal, such as [Pull out] from a detail to the full scene.
- Social video templates built around a single move, for example a slow [Push in] on a portrait or a [Tracking shot] behind someone walking.
Inputs
A single prompt field holds both the scene and the camera directions, up to 2,000 characters. Write each command in square brackets, using these names:
- Truck left, Truck right
- Pan left, Pan right
- Push in, Pull out
- Pedestal up, Pedestal down
- Tilt up, Tilt down
- Zoom in, Zoom out
- Shake
- Tracking shot
- Static shot
Commands grouped inside one bracket, such as [Truck left, Pan right, Zoom in], happen at the same time. Commands placed at different points in the text happen in that order.
Outputs
Each run returns one silent H.264 MP4 at 1280x720. The sample runs 5.6 seconds at 25 frames per second, close to the 6-second length MiniMax gives as this model's default.
Recommended settings
- Put no more than 3 movements inside one bracket. That is the limit MiniMax recommends.
- Spell the commands exactly as listed. According to MiniMax, plain-language camera directions also work, but the listed commands are followed more accurately.
- After each command, describe what should be on screen next, the way the default prompt follows [Shake] with the change in the woman's expression.
Limitations
- Every clip is 720p landscape and close to 6 seconds long, with no setting on this page to change either.
- There is no image input here. The prompt optimizer, which MiniMax turns on by default, has no switch on this page, so your prompt may be rewritten before generation.
Safety & compliance
If MiniMax flags a prompt as sensitive, generation is refused and no clip comes back.
API quick start
Run text-to-video-hailuo-01-director with a single API call.
{
"prompt": "[Push in]Close up of a tense woman looks …"
}curl -X POST "https://api.wiro.ai/v1/Run/minimax/text-to-video-hailuo-01-director" \
-H "Content-Type: application/json" \
-H "x-api-key: YOUR_WIRO_API_KEY" \
--data-binary @- <<'JSON'
{
"prompt": "[Push in]Close up of a tense woman looks …"
}
JSONOther models you might like
Explore similar video generation models.