Communitygithub.com

BennyKok/omg.dev

Generate images and videos from a prompt on an omg.dev Computer with omg_generate_image and omg_generate_video. Every call spends the user's omg credits. Use when the task needs a new picture, icon, illustration, or short clip, not for editing a file that already exists.

What is omg.dev?

omg.dev is a Claude Code agent skill that generate images and videos from a prompt on an omg.dev Computer with omg_generate_image and omg_generate_video. Every call spends the user's omg credits. Use when the task needs a new picture, icon, illustration, or short clip, not for editing a file that already exists.

Works with~Claude Code~Codex CLI~Cursor
npx skills add https://github.com/BennyKok/omg.dev/tree/HEAD/agents/skills/media-generation

Ask in your favorite AI

Open a new chat with this agent skill pre-loaded.

Documentation

Generate images and videos

These tools spend the user's omg credits. 1 credit is 1 USD. Treat every call as a purchase.

When to use

  • Use them when the task needs new visual media: a hero image, an icon, a logo, an illustration, a short clip.
  • Do not use them for a screenshot, a chart, or anything you can draw with code or SVG.
  • Do not generate many variations "to be safe". Make one, show it, and ask if the user wants another.

Rules

  1. Leave model out unless the request needs a specific look. With no model the router picks the cheapest curated model for the task ("auto") and names it in the result.

  2. Each result has model and costUsd. Tell the user both in your reply, for example "Generated with flux-schnell for $0.003."

  3. A call is refused above the per-call cap ($1.00 default) or when the daily cap ($25.00 default, UTC day) is used up. The error says how much is left. Do not work around the cap. Tell the user.

  4. If the result has pending: true, call omg_media_job with the jobId. Do not generate again; that charges twice.

  5. If the error says the credits are used up, tell the user to add credits at omg.dev. Do not retry.

  6. Show images with omg_display_image. Show videos with omg_display_video.

  7. A video for the iPhone app must be under 6 MB, H.264, with faststart. The tools already remux downloads to faststart. Check the size. If it is larger, re-encode and keep the original:

    ffmpeg -i in.mp4 -vf scale=480:-2 -c:v libx264 -profile:v main -crf 27 -maxrate 1800k -bufsize 3600k -c:a aac -b:a 96k -movflags +faststart out.mp4
    

Choosing a model

  • Usually pass no model. The router picks the cheapest suitable model for the task and names it in the result.
  • Pass model only when the request needs a specific look, for example readable text, SVG, or best-quality video.
  • omg_media_models lists the top models with prices. omg_media_models {all: true, q: "kling"} searches the full list.
  • A model outside the top list is priced before it runs. If it cannot be priced, the call is refused; pick another.

Video takes 30 seconds to 5 minutes. Images take 3 to 20 seconds.

Fields

  • Friendly fields map to provider input: prompt, aspectRatio to aspect_ratio, durationSeconds to duration, resolution to resolution.
  • Put any other provider field in input, for example {"generate_audio": false} or {"quality": "high"}.
  • Files save to ~/omg-media/<date>/<jobId>-<n>.<ext> unless you pass outputPath.

3D models

Use omg_generate_3d to turn one image of an object into a GLB file. Give imageUrl, or imagePath for a local image under 600 KB. Use one object on a plain background.

  • tripo3d/h3.1/image-to-3d is the default. It costs $0.30. Use {face_limit: 8000} for a game-ready mesh.
  • fal-ai/hyper3d/rodin/v2.5/fast costs $0.10. Use it for quick drafts.
  • fal-ai/trellis-2 costs $0.25 to $0.35, by resolution.
  • fal-ai/hunyuan3d-v3/image-to-3d costs $0.375. {generate_type: "LowPoly"} costs $0.45.
  • PBR, extra views, and detailed geometry cost more. The quote includes them.

Individual skills in this repo

This repo contains 4 individual skills — each has its own dedicated page.

Related Skills