About this skill

Problem

When you already have a first frame and a tail frame, direct video generation can still produce jumps, object drift, or broken motion in between. This skill targets that keyframe-to-transition case by calling Jimeng's first and tail frame model to generate the connecting motion from a prompt and the two boundary frames. It is useful for turning storyboard keyframes, shot open/close states, or the start and end of a continuous action into a playable video segment.

How It Works

Core capability: For jimeng i2v first tail style requests, it sends the first frame, tail frame, prompt, and generation parameters to the dLazy hosted API for inference.
Main flow: The skill uses the dLazy CLI to run dlazy jimeng-i2v-first-tail; local image, video, or audio paths are uploaded to dLazy media storage so the model can read them; the API returns output URLs hosted on files.dlazy.com.
Pipeline support: Flags accept pipe references such as -, @N, @N.path, @*, and @stdin, which helps chain upstream outputs into the current task.
Async tasks: With --no-wait, the response returns a generateId and status; use dlazy status --wait to poll the result.

Boundaries And Notes

It requires a dLazy API key and available account balance; insufficient_balance should prompt the user to recharge.
Missing or invalid credentials return unauthorized; the user needs to obtain and save a valid API key.
Input files leave the local environment and are uploaded to dLazy media storage, so review service terms before processing sensitive assets.
Generation can fail if the prompt violates safety policy, if local files are missing, if required parameters are absent, or if the backend returns an error.

Screenshots

Use Cases

When producing storyboards, use existing first and tail keyframes to generate the connecting motion.
In short-video post-production, connect two visual states into a coherent segment instead of a hard cut.
In multi-step pipelines, pass upstream first and tail outputs into Jimeng using pipe references like `@N`.
In async batch jobs, submit first-tail video tasks with `--no-wait` and poll status by task ID.

Best For

AIGC video editors who connect storyboard shots and need first-tail frame transitions.
Agent developers building multi-step video pipelines who call the dLazy CLI for first-tail tasks.
Creative team engineers using dLazy cloud inference who manage API keys, balance, and async task status.
Video creators making product demos who need to connect opening and closing shots into continuous motion.

Related Skills

ComfyUI Interactive Image Generator

@user_99514c20/comfyui-interactive-image-generator

A ComfyUI image-generation skill that uses a five-step dialog to collect prompts, prefill templates, confirm parameters, submit API jobs, and return results.

Design & Media 2026.08.29 ↓ 606