kapi exec media-refine

$ kapi media-refine [files...]

Re-read low-confidence OCR/ASR lines with a configurable multimodal LLM

Commandexec media-refine

Try it (demo)

Demo mode — illustrative output from a built-in stub, not a real model. Install the CLI to run with your own API key.

$ kapi exec media-refine

Flags

Flag	Type	Default	Description
--api-key	string		API key for the AI provider
--concurrency, -j	int	0	max parallel files (0 = auto)
--credential	string		saved credential name to use (see 'kapi credentials list')
--encoding, -e	string	UTF-8	input file encoding
--fail-on-unknown	bool	false	exit with error if any file cannot be processed (default: skip with warning)
--format, -f	string		override input format detection
--json	bool	false	output results as JSON
--map, -m	stringArray		map glob pattern to format (e.g. '*.docx=okf_openxml:test')
--model	string		Multimodal model name
--no-warn	bool	false	suppress warnings for skipped files
--output, -o	string		output path template (variables: {dir}, {name}, {ext}, {lang})
--output-dir	string		write outputs under DIR/{lang}/ (default: beside the input, mirroring its locale layout)
--pack	bool	false	when transforming a .kpz, also eject the result to the .kpz (auto-pack)
--parallel-blocks	int	0	fan out block processing across N goroutines (0 = off)
--progress, -p	bool	false	show progress bar
--provider	string	anthropic	Multimodal AI provider (Anthropic, OpenAI, Gemini, Azure OpenAI, Claude Code, Ollama, Demo (illustrative))
--source-lang	string	en	source language (e.g. en, en-US)
--strict	bool	false	alias for --fail-on-unknown
--target-lang	string		target language (e.g. fr, de-DE)
--threshold	float64	0.85	Re-read extracted lines whose confidence is below this
--trace	string		write flow trace JSON to file (for flow visualization)