Guide · Agent skill
How to use the fal.ai Media Generation agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The fal-ai-media skill gives an AI agent the instructions it needs to generate images, videos and audio by calling fal.ai's hosted models through an MCP server, which is a connector that lets the agent use outside tools. This guide is for people who have not used agent skills before and want a working setup. After following it, you will know what the skill does, how to install it, how to phrase requests so the agent uses it, and what to check for cost and safety.
fal.ai Media Generationaffaan-m/ECC · MIT Download the skill Quick start
- 1Create a fal.ai account. Sign up at fal.ai and create an API key for the MCP server.
- 2Add the MCP entry. Add the fal-ai block to ~/.claude.json with your FAL_KEY in the env section.
- 3Download and unzip the skill. Download the ZIP from this page and extract it so the skill folder contains SKILL.md.
- 4Place the skill folder. Put it in ~/.claude/skills/ for personal use or .claude/skills/ inside a project, or upload it under Settings → Capabilities → Skills in the Claude apps.
- 5Make a request. Ask for an image, video or audio in plain language, including the subject and format.
- 6Check the estimate. Ask the agent to run estimate_cost before expensive video generations.
What the fal-ai-media skill makes the agent do
An agent skill is a folder of instructions that an AI agent loads when a task matches the skill's description. This one covers media generation: images, video and audio. The SKILL.md file lists the models it knows about, including Nano Banana 2 and Nano Banana Pro for images, Seedance 1.0 Pro, Kling Video v3 Pro and Veo 3 for video, CSM-1B for text-to-speech, and ThinkSound for video-to-audio.
- Text-to-image, plus image editing with an uploaded source image
- Text-to-video and image-to-video
- Text-to-speech with CSM-1B and video-to-audio with ThinkSound
- Optional direct calls to ElevenLabs and VideoDB for voice and music
Install the fal.ai MCP server before using the skill
The skill does not work on its own. It needs the fal.ai MCP server configured in your ~/.claude.json file, with your fal.ai API key set as FAL_KEY. Get an API key at fal.ai. The server is started with npx, so Node.js must be installed.
Add this entry under your MCP configuration in ~/.claude.json, replacing the placeholder with your own key:
- Requires a fal.ai account and API key
- Requires Node.js with npx
- Storing FAL_KEY in ~/.claude.json means the key sits in a plain config file
"fal-ai": {
"command": "npx",
"args": ["-y", "fal-ai-mcp-server"],
"env": { "FAL_KEY": "YOUR_FAL_KEY_HERE" }
}Install the skill from the ZIP or the Claude apps
Download the ZIP for this skill from this page and unzip it. For a personal install, place the skill folder in ~/.claude/skills/. For a single project, place it in the project's .claude/skills/ folder. If you use the Claude apps, you can instead upload the skill under Settings → Capabilities → Skills.
The skill's package includes agents/openai.yaml and SKILL.md. SKILL.md is the file that contains the instructions.
- Personal install: ~/.claude/skills/
- Project install: .claude/skills/ inside the project
- Claude apps: Settings → Capabilities → Skills
Trigger the skill with natural requests
You do not need to name the skill. The skill activates when your request matches its description. The SKILL.md file lists phrases such as "generate image", "create video", "text to speech" and "make a thumbnail" as triggers, along with general requests for speech, music or sound effects.
Be specific about the output type and the subject. Mention the format or length when it matters, since the skill uses parameters such as aspect ratio and duration.
- "Generate an image of a futuristic cityscape at sunset in cyberpunk style."
- "Make a thumbnail for my video about sourdough, 16:9."
- "Create a 5 second video of a drone flyover of a mountain lake at golden hour."
- "Text to speech: read this welcome line in a conversational voice."
Use the tools and parameters the skill describes
The skill tells the agent to use fal.ai MCP tools: search, find, generate, result, status, cancel, estimate_cost, models and upload. Common image parameters include image_size (square, portrait_4_3, landscape_16_9, portrait_16_9, landscape_4_3), num_images (1 to 4), seed and guidance_scale (1 to 20). Video parameters include duration ("5s" or "10s"), aspect_ratio (16:9, 9:16 or 1:1) and image_url for image-to-video.
You do not write these calls yourself in normal use. Reading them helps you judge what the agent is about to run.
generate(
model_name: "fal-ai/nano-banana-2",
input: {
"prompt": "a futuristic cityscape at sunset, cyberpunk style",
"image_size": "landscape_16_9",
"num_images": 1,
"seed": 42
}
)Check cost and understand the optional API examples
The skill tells the agent to check estimated cost with estimate_cost before running a generation, and the SKILL.md tips say to check it before expensive video generations. Start with Nano Banana 2 for drafts, then switch to Nano Banana Pro for final images. The skill also contains direct API examples for ElevenLabs (voice, needs ELEVENLABS_API_KEY) and VideoDB generative audio (needs VideoDB configured). These bypass MCP and need their own keys.
- Use seed to reproduce results while you refine a prompt
- Image-to-video gives more controlled results than pure text-to-video
- Keep video prompts descriptive but concise, focused on motion and scene
Safety notes: scripts, network and rights
This skill runs scripts and uses the network. The MCP server is started through npx with fal-ai-mcp-server and sends your prompts and any uploaded files to fal.ai. The ElevenLabs and VideoDB examples send requests to their own services. Review the configuration before you add your key.
The skill file does not describe the licensing terms of the models it calls. Outputs may resemble existing works, so do not rely on them where originality or clearance must be guaranteed.
- Keep FAL_KEY out of shared or synced config files where possible
- Generations cost money; check estimates first
- Verify rights before commercial use of outputs
What people use it for
- Generating a 16:9 thumbnail or product photo draft with Nano Banana 2 or Nano Banana Pro
- Turning a still image into a short clip with image-to-video using Seedance 1.0 Pro
- Producing a narration line with CSM-1B text-to-speech
- Adding matching ambient audio to a video with ThinkSound
Frequently asked questions
Is the fal-ai-media skill free?
The skill is released under the MIT license. fal.ai generations are billed through your fal.ai account, and the skill directs the agent to estimate cost before running them, so check your account's pricing before generating at volume.
What do I need to install the fal.ai media skill?
You need a fal.ai account and API key, Node.js with npx, and the fal-ai MCP server entry in ~/.claude.json with FAL_KEY set. ElevenLabs and VideoDB are optional and only needed for their examples.
Where do I put the skill folder?
Place it in ~/.claude/skills/ for a personal install or in .claude/skills/ inside a project. In the Claude apps, you can upload it under Settings → Capabilities → Skills.
Why isn't the skill generating anything?
The most likely cause is that the fal-ai MCP server is not configured in ~/.claude.json or FAL_KEY is missing or wrong. Confirm the entry matches the example in this guide, then ask the agent to use the MCP tools again.
Which model should I use for images?
The skill describes Nano Banana 2 as fast and suited to drafts and editing, and Nano Banana Pro as higher fidelity for production images, realism and typography. Use the faster one to iterate and the Pro model for finals.
Does the skill use ElevenLabs or VideoDB automatically?
No. The ElevenLabs and VideoDB sections are direct API examples that need their own keys or configuration. Only the fal.ai MCP tools run through the server.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use this skill if you want an agent to produce quick image, video and audio drafts through fal.ai and you are comfortable storing an API key in a config file and watching generation costs. Pick something else if your outputs must be rights-cleared, if your budget has no room for cost checks, or if you will not keep a key in ~/.claude.json.
Download the skill