Direct a continuous world in real time with image references and text prompts.

minimax/h3-max/reference-director

Direct a continuous world in real time with image references and text prompts.

Learn more about H3 Max

Inference

Commercial use

Schema AsyncAPI

LLMs

Playground API

Input

Prompt*

Additional Settings

Customize your input with more control.

Result

Idle

What would you like to do next?

minimax/h3-max/reference-director

Schema

Input

Prompt*

Additional Settings

Customize your input with more control.

Result

Idle

What would you like to do next?

Input

Checking image dimensions…

Start with an example

The Last FerryCinema collection / Cinema

The Parcel Takes the CanalCinema collection / Cinema

The Shadow RaceCinema collection / Cinema

The Umbrella TrainCinema collection / Cinema

Following the SwellCinema collection / Cinema

The Cloud ShortcutCinema collection / Cinema

Jim & Silver — The Set That Would Not EndCinema collection / Cinema

The Rooftop RelayCinema collection / Cinema

The House Is Waiting for SomeoneCinema collection / Cinema

Prompt*

Resolution

Aspect Ratio

Reference images

Reference image 1

Name

Source URL*

Reference image 2

Name

Source URL*

Script

Reference audio

Pin audio

Memory

Number of prior segment prompts retained as context for future prompt expansion.

Audio Bitrate

Session audio target in bits/s: 96000, 128000, or 192000. Explicit values use Opus audio mode on direct WebRTC; LiveKit uses its native mode with the same bitrate target. Null preserves transport defaults (WebRTC: 96000, voip). Immutable; reconnect to compare.

Seed

Evaluations

MXFP8 attention

Use MXFP8 attention for dense generation steps.

Forced cut max segments

Most segments in one unbroken continuation run before a forced hard cut opens a fresh shot (Ref2VA burn-in mitigation). 0 disables forced cuts; omitted keeps the deployment default. Pinned for the whole session.

Dialogue words per second

Spoken words per second of new screen time the planner budgets dialogue against. The default keeps delivery calm; raise it for deliberately fast talkers, lower it for sparser dialogue.

Result

Idle

Sessions cost $0.00 per compute second of video generated.

Need sessions longer than 15 minutes?

We’re gradually enabling longer Director sessions for approved use cases.

Input

Prompt*

Additional Settings

Customize your input with more control.

Result

Idle

What would you like to do next?

Input

Checking image dimensions…

Start with an example

The Last FerryCinema collection / Cinema

The Parcel Takes the CanalCinema collection / Cinema

The Shadow RaceCinema collection / Cinema

The Umbrella TrainCinema collection / Cinema

Following the SwellCinema collection / Cinema

The Cloud ShortcutCinema collection / Cinema

Jim & Silver — The Set That Would Not EndCinema collection / Cinema

The Rooftop RelayCinema collection / Cinema

The House Is Waiting for SomeoneCinema collection / Cinema

Prompt*

Resolution

Aspect Ratio

Reference images

Reference image 1

Name

Source URL*

Reference image 2

Name

Source URL*

Script

Reference audio

Pin audio

Memory

Number of prior segment prompts retained as context for future prompt expansion.

Audio Bitrate

Seed

Evaluations

MXFP8 attention

Use MXFP8 attention for dense generation steps.

Forced cut max segments

Dialogue words per second

Spoken words per second of new screen time the planner budgets dialogue against. The default keeps delivery calm; raise it for deliberately fast talkers, lower it for sparser dialogue.

Result

Idle

Sessions cost $0.00 per compute second of video generated.

Need sessions longer than 15 minutes?

We’re gradually enabling longer Director sessions for approved use cases.