Generate AI images and videos via Modellix CLI

This skill teaches agents to use Modellix, a Model-as-a-Service platform for AI image and video generation, through its official modellix-cli workflow: authenticate, check environment health with doctor, submit a generation with model run --wait, then persist the result with task download. It covers text-to-image and text-to-video generation and calling Modellix models through a unified API/CLI, and applies when a user mentions Modellix or specific providers routed through it, like Seedream, Seedance, or Nano Banana.

The workflow runs in five steps: authenticate via MODELLIX_API_KEY or modellix-cli auth login, run modellix-cli doctor --json to check the environment, fall back to sensible default models when none is specified (text-to-image: google/nano-banana-2-lite; text-to-video: bytedance/seedance-2.0-mini-t2v), submit with modellix-cli model run --wait --json, and persist outputs with modellix-cli task download. A representative text-to-image call:

modellix-cli model run \
  --model-slug google/nano-banana-2-lite \
  --body '{"prompt":"A cinematic sunset over a futuristic city"}' \
  --wait --timeout 5m --json

Use it for generating or editing images and video from text or existing images through Modellix's models. It builds in real cost and safety discipline rather than assuming it away: before any paid submission it discloses the provider, model, prompt or source media that will leave the machine, expected cost, and output path, and requires explicit user approval - it does not blindly retry a paid submission after an unknown outcome, instead checking task history first. It prefers session-scoped API key use, running auth login for persistent local credential storage only when the user has approved that. It also confirms the destination and overwrite policy before task download and never replaces an existing file without explicit approval.

Input is a text prompt (and optionally source media) plus a model slug, with request schemas fetched from model describe's docs_url or Modellix's own llms.txt documentation. Output is the generated image or video, downloaded and persisted locally via task download.

It integrates with the modellix-cli tool and Modellix's hosted API (api.modellix.ai) and CDN, which process prompts and any uploaded source media remotely - requiring a valid API key and never logging secrets.

Developers and agents generating AI images or video through Modellix who need a disclosed, approval-gated workflow rather than blind paid API calls - output still requires human review for quality, rights, privacy, and policy compliance before publication, since availability, pricing, quotas, and moderation are all controlled by Modellix and subject to change.