Dreamina CLI

Problem

When integrating Dreamina, also known as 即梦, into scripts, agents, or automation, the pain points are concrete: login sessions are not obvious to reuse, image and video commands have different flags, async jobs return a submit_id before the result exists, and some models may require one-time web authorization. dreamina-cli treats the CLI as the primary interface for Dreamina image and video generation, plus account and session operations.

How it works

The workflow starts with dreamina -h, then inspects each relevant subcommand with -h before running real work. The skill emphasizes reusing existing login state unless the user asks for login or relogin, warning before credit-consuming commands, and choosing the right operation: user_credit for budget, image or video for first outputs, image2video for a single image to video, multiframe2video for multiple images into one coherent story, and multimodal2video for richer image/video/audio references, including the seedance2.0 family when appropriate. For async generation, a submission is successful only when a submit_id is present and gen_status is querying or success; if it fails, inspect fail_reason. Follow-up should use query_result --submit_id= for a known job and list_task for saved history. Model selection must not be assumed from the skill: check the subcommand help for available models, duration, ratio, resolution, and other constraints.

Limits and cautions

This is useful for command-driven workflows where help inspection, paid submission, and async follow-up should be kept separate. It is not a substitute for checking each subcommand's supported combinations, because different commands may expose different models, ratios, durations, and resolutions. If the CLI returns AigcComplianceConfirmationRequired, the user should complete the web-side confirmation before retrying.