Jimeng Image-to-Video CLI

Problem

Jimeng image-to-video CLI work often fails when the wrong command is chosen, required parameters are missing, resolution or model support is mismatched, or async results are unclear. image2video, frames2video, multiframe2video, and multimodal2video have different key inputs: single image tasks need --image and --prompt, first/last frame tasks need --first and --last, storyboard tasks need --images, and multimodal references need at least one image or video. This skill turns routing, validation, submission, and result polling into a concrete workflow.

Core capabilities and key steps

• Intent routing: choose the command based on single-image animation, first/last frame transition, 2–20 image storyboards, or image/video/audio references.
• Parameter validation: check model, duration, and resolution with the matching dreamina -h; all modes require an explicit --video_resolution using lowercase 720p, 1080p, or 4k.
• Hard constraints: multiframe2video uses a fixed model and accepts only 720p/1080p; N images need N−1 transition entries, with segment duration between 1 and 8 seconds.
• Async loop: check credits and explain consumption impact, save submit_id, keep calling query_result until success or fail, report fail_reason, and prompt users to complete Web authorization when compliance confirmation is required.

Boundaries

Use it when a reference image or video exists. It is not for pure text video, image generation, or prompt polishing. multimodal2video supports only the Seedance 2.0 family, and audio cannot be the only input. image2video and frames2video have different model support, so follow the version matrix.