miketromba/ploof

This is a command-line tool for generating images, audio, and video files using AI services, designed to be operated by AI coding assistants rather than typed manually. The primary use pattern is to describe what you want to a coding agent like Claude Code, Cursor, or Codex, and let that agent install the tool, read its built-in documentation, and produce the files on your behalf. When run by a coding agent, the workflow starts with the agent installing ploof, then running a command called ploof learn, which prints complete documentation for the installed version in a format agents can parse reliably. The agent then authenticates with the relevant provider and runs the appropriate generation commands. Every output file gets a companion JSON file alongside it that records the exact prompt and parameters used, so results are reproducible. The tool supports images, video, and audio across two AI providers. For OpenAI it covers image generation, editing, and variations, video creation and editing, text-to-speech, transcription, and translation. For fal.ai, a marketplace of AI models, it supports all three media types and can also run any available model endpoint directly. Multiple API keys per provider can be stored in named profiles, which is useful when switching between personal and work credentials or in automated pipelines where keys come from environment variables. For batch work, assets can be declared in a YAML file with dependencies between them, then generated in parallel with one command. Output throughout the tool is structured as JSON or line-delimited JSON, making it straightforward to pipe into other scripts or for agents to parse results automatically.