Jimeng Text-to-Video CLI

Problem to solve

When an agent calls Jimeng text-to-video via CLI, failures often come from parameter and workflow mistakes rather than prompt quality: missing --video_resolution, using uppercase resolution tokens, selecting unsupported models or durations, and resubmitting while the task is still querying. This skill turns the Jimeng CLI text-to-video path into a repeatable execution workflow.

How it works

It first runs dreamina text2video -h, treating the help output as the source of truth for parameters, then follows three steps:
- Validate: check model, duration, ratio, and resolution compatibility; public text-to-video models are limited to the Seedance 2.0 family.
- Submit: explain credit consumption, prefer asynchronous submission, and save the submit_id.
- Close the loop: keep calling query_result until success or fail, report fail_reason when needed, and direct the user to Web authorization if AigcComplianceConfirmationRequired appears.

Key constraints include lowercase 720p, 1080p, and 4k; 1080p and 4k are available only for seedance2.0_vip; duration is 4–15 seconds, with seedance2.0fast as the default model. If polling times out but the status remains querying, do not immediately resubmit. This skill does not cover image-to-video, storyboards, or prompt creation.