mediagen: image and video generation for agents
mediagen
A skill and CLI for image and video generation through Gemini, OpenAI and Kie AI, with provider choice and AI-content marking
- 50 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- Requests go to providers under your keys and spend a paid quota
- API keys are stored in a config file or the environment on your machine
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/mediagenDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Requests go to providers under your keys and spend a paid quota
- API keys are stored in a config file or the environment on your machine
This is third-party code. Review the repository files before installing.
What it does
The mediagen skill teaches the agent to call the mediagen CLI through npx -y mediagen with the --json flag and receive exactly one JSON object instead of prose. The agent writes a detailed prompt itself, picks the provider and model, and reads the file path from the response. If you name a model, aspect ratio, size or filename, it uses them as given and decides the rest or takes it from settings. There are three providers. Gemini does images up to 4K, video and editing, OpenAI covers the gpt-image family and DALL·E, and Kie AI opens about thirty third-party models including Flux, Imagen and Grok. mediagen models shows the provider order and what a request would get. The agent never asks for an API key in chat: on a configuration error it sends you to run mediagen init in your own terminal. mediagen mark puts an AI-content label on a finished file: a machine-readable IPTC/XMP marker and, optionally, a visible AI GENERATED sign from the European Commission's code of practice. Marking is always a separate second step, so the sign can go where the subject is not.
Who it is for. Designers, marketers and developers who need banners, illustrations and short clips straight from Claude Code or Codex without leaving for web interfaces.
Good fit when
- A landing page needs a wide 21:9 banner and the agent must pick a provider that can produce that shape
- A mockup needs four 4K illustration variants with one filename and output directory
- A finished image is published in the EU and needs both a machine-readable and a visible AI label
- You have a photo to change from a description, for example making the sky stormy
- You need a six-second clip from a description; only Gemini can do video
Not a fit when
- You need precise photorealistic output with no human review: judge the prompt and result by eye before release
- None of the providers has an API key: generation cannot start without at least one
- You need video on OpenAI or Kie AI: only Gemini makes video
- A stock image from the brand's existing library solves the task
Example request
Make a hero image for the landing page in 21:9, moody industrial style, and put a visible AI label in the top right corner.Limitations
You need a key for at least one provider and Node.js 20.11 or newer. Generation is billed to your key, and the README gives no provider pricing. Only Gemini makes video, and it takes minutes. OpenAI cannot produce 16:9, its widest image is 1536 by 1024, and a 16:9 request is refused by name.
Kie AI has no cheap key check, and the doctor report says so. Marking a JPEG or WebP means re-encoding it, a second lossy pass. Access to the providers from Russia was not checked for this card.
How to disable. Delete the skill folder from the agent's skills directory. Keys saved in settings are removed with mediagen config edit, and generated files stay in the ./output folder.
FAQ
How do I install the mediagen skill?
With npx -y skills add Cripacx/mediagen --skill mediagen. The CLI needs no separate install: the skill runs it through npx and caches it on first use. Then run npx -y mediagen init once to enter keys.
Do I give the agent my API key?
No. The CLI deliberately has no flag that takes a key as an argument, since it would land in shell history and the process list. The init wizard takes keys in your terminal without echo, and settings resolve from the environment, then .env, then the config file.
Which provider fits video and wide images?
Only Gemini makes video, and it also handles 21:9. OpenAI cannot do 16:9; Kie AI has about thirty models for images. Without --provider, a request goes to the first provider in order that has a key and can do the job.
What does mediagen mark do?
It writes IPTC/XMP DigitalSourceType metadata and, with --visible-label, composites an AI GENERATED or AI MODIFIED sign. The result is saved as a new file with a labelled suffix and the original stays untouched unless --in-place is set.
Does mediagen work without an agent?
Yes, it is a standalone CLI. npx -y mediagen image, video, models and mark work from a terminal. Install the package globally if you call it often. Hosts with no shell get an MCP mode: npx -y mediagen mcp.
Discussion
No questions or reviews yet.
Related
SkillsA set of skills for agents to generate and edit images, video and audio through the muapi.ai service and muapi-cli
A CLI for fal.ai: an agent searches, runs and downloads results from 1200+ image, video and audio models with shell commands.
A skill that writes prompts for video, image and music generation on 14 platforms and can submit them in the browser
Image generation and editing from any agent: the imagegen-agent CLI and a skill with the Codex prompt library for compatible APIs
Generate, edit and deliver images through your API providers or the Codex ChatGPT route, with a canvas and version history.
Two skills for generative video and images: video writes Seedance, Kling and Veo prompts by dramaturgy rules, image picks the model and its syntax
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| mediagen: image and video generation for agents (this card) | no | 50 | ||
| claude-image-gen | no | 83 | generation tied to Claude with no choice among three providers | |
| MCP Image Generator | no | key is optional | 190 | an image MCP server with no AI-content marking step |
README in short
The README shows two scenarios: the user speaks roughly and the agent chooses the parameters, or names everything exactly and the agent decides nothing. Then come prerequisites, configuration and provider ordering, a provider table, a section on marking under EU rules, and use without an agent.
README badge
Original SKILL.md
--- name: mediagen description: Generate images and video from text prompts via the mediagen CLI, across Gemini, OpenAI and Kie AI. Use when asked to create, generate, render, or edit an image or a video — a logo, an illustration, a mockup, a photo, a banner, a clip — or to mark existing media as AI-generated. Also covers choosing a provider or model for a generation task. --- # mediagen Generate images and video from text prompts across several providers. ## How to call it Use **`npx -y mediagen`** with `--json`. Always `--json`: it puts exactly one object on stdout and nothing else, so you never have to parse prose. ```bash npx -y mediagen image "<prompt>" --json npx -y mediagen video "<prompt>" --json ``` `npx` needs no installation and runs an existing one if the user has it, so this is the right form either way — installing the skill does not install the CLI, and most users who have the skill will not have it on their PATH. **Always pass a subcommand.** With no arguments the tool prints its help and exits; it does nothing useful, but it will not hang. Do not run `mediagen init` yourself — it needs a terminal and refuses without one. It is something to ask the user to run. If you have no shell at all, the same pipeline is available over MCP as the `generate_media` tool, with `list_models` and `check_configuration` alongside it — a host spawns it with `npx -y mediagen mcp`. The CLI is the primary path; MCP is the fallback. Video takes minutes and only Gemini does it. Progress goes to stderr; the one JSON object still lands on stdout at the end. ## Write the prompt yourself **This is the part that decides whether the result is any good, and it is your job.** mediagen sends your prompt to the provider exactly as written — it does