Private media
Inputs and results use private object storage and short-lived viewing links. Files are retained for 30 days.
Text to video
Start with language. Let motion carry the idea.
Text mode is suited to concept shots, atmospheric transitions, product mood studies, and scenes where exact boundary frames are not required.
MiniMax H3
Create a new video
Private by default
Sample directionFluid form / controlled orbit
15-second output
One workspace
Choose the control the shot needs
Each mode produces the same private task record, so changing creative direction does not split your files or credits across separate tools.
Task safeguards
Generation that keeps its accounting straight
One charge per task
Explicit upstream failures can retry three times without another user charge. Final failure releases the reservation.
Durable task state
Queued, running, reconciling, and finalizing states survive navigation and worker restarts.
Questions
MiniMax H3 on VideoCurrent
Which MiniMax H3 modes are available?
VideoCurrent supports text to video, first frame, last frame, and combined first-and-last-frame generation.
Are generated videos public?
No. Inputs and results are private by default. A result is public only when you create a revocable share link.
How long are files retained?
Input images and generated videos are retained for 30 days. Task and billing records remain for account and audit history.
What happens when a generation fails?
VideoCurrent can retry an explicitly failed generation up to three times at no extra charge. A final failure releases the credits reserved for that task.