CLI reference

Command shape

quickedit INPUT_PATHS... [OPTIONS]

Configuration priority is CLI option, environment variable, JSON config file, then built-in default. Environment variables use the QUICKEDIT_ prefix.

Input and output

Argument or option	Default	Purpose
INPUT_PATHS	required	One or more input video files. Multiple inputs run as a batch.
-o, --output PATH	<input>_edited.ext	Output video path for one input, or output directory for a batch.
--overwrite	off	Replace an existing output video intentionally.
--dry-run	off	Analyze, build the timeline, and write sidecars without rendering.

QuickEdit refuses to replace an existing output unless --overwrite is set. In batch mode, a failed file is reported and the remaining inputs continue.

Configuration and environment

Option	Default	Purpose
--config FILE	none	Read a JSON config file. Keys use underscores, such as whisper_model.

Every CLI option can also be set as a QUICKEDIT_ environment variable by uppercasing the option name and replacing dashes with underscores. For example, QUICKEDIT_WHISPER_MODEL=small sets the same value as --whisper-model small.

Configuration file

Pass --config settings.json. Keys use underscores, for example:

{
  "whisper_model": "small",
  "motion_frame_skip": 2,
  "motion_backend": "ffmpeg",
  "motion_workers": 4,
  "subtitle_style": "simple",
  "use_llm": false,
  "preset": "fast"
}

Speech and transcription

Option	Default	Purpose
--whisper-model TEXT	base	Whisper model size. Common values are tiny, base, small, medium, and large-v3.
--device TEXT	cpu	Device for faster-whisper, usually cpu or cuda.
--whisper-compute-type VALUE	auto	Compute type: int8, float16, float32, or int8_float16. If omitted, QuickEdit uses int8 on CPU and float16 on CUDA.
--language TEXT	auto	Language code such as en. Leave unset to auto-detect.
--vad-threshold FLOAT	0.5	Voice activity sensitivity from 0.0 to 1.0. Lower values are more sensitive.

Transcription runs when subtitles, LLM editing, or a words combine detector needs word timestamps. Speech detection uses VAD and does not require subtitles.

Motion analysis

Option	Default	Purpose
--motion-threshold FLOAT	0.02	Fraction of changed pixels required for a frame to count as active.
--motion-pixel-threshold INTEGER	10	Pixel brightness change from 0 to 255 required before a pixel counts as changed.
--motion-frame-skip INTEGER	1	Analyze every Nth frame. Higher values are faster but can miss short changes.
--motion-backend VALUE	ffmpeg	Motion backend: ffmpeg, opencv, or opencv-parallel.
--motion-workers INTEGER	4	Worker count for --motion-backend opencv-parallel.

Motion analysis is skipped when --combine does not reference motion and LLM editing is disabled.

Editing logic

Option	Default	Purpose
--combine TEXT	or:speech,motion	Combine detector arrays. Examples: speech, or:speech,motion, and:speech,motion, xor:speech,motion, not:motion.
--margin FLOAT	0.2	Add padding before and after kept regions, in seconds.
--minclip FLOAT	0.1	Remove keep segments shorter than this many seconds.
--mincut FLOAT	0.2	Fill cut gaps shorter than this many seconds.

The available detector names are speech, motion, and words. words uses transcribed word timestamps, so it requires transcription.

Optional LLM editing

Option	Default	Purpose
--llm	off	Enable semantic LLM editing. Transcript data and prompt context are sent to the selected provider.
--no-llm	on	Disable semantic LLM editing. This is the default.
--llm-model TEXT	claude-sonnet-4-20250514	LLM model name. claude-* uses Anthropic; gemini-* uses Google.
--prompt TEXT	none	Inline custom editing instructions for the LLM.
--prompt-template VALUE	none	Built-in prompt template: dsa, tutorial, or lecture.
--prompt-file PATH	none	Read custom editing instructions from a file.
--confidence FLOAT	0.7	Minimum confidence for LLM-suggested cuts from 0.0 to 1.0.
--silent-segment-min-duration FLOAT	0.5	Minimum silence duration, in seconds, before a silent segment is classified for LLM context.

For Anthropic models, set ANTHROPIC_API_KEY. For Gemini models, set GOOGLE_API_KEY or GEMINI_API_KEY.

Subtitles

Option	Default	Purpose
--subtitle-style VALUE	fancy	Subtitle style: fancy for ASS word highlighting, simple for SRT, or none.
--subtitle-font TEXT	Arial	Font name used for ASS subtitles.
--subtitle-size INTEGER	20	Font size used for ASS subtitles.
--subtitle-silence-gap FLOAT	0.7	Silence gap, in seconds, that starts a new subtitle group.

Subtitle sidecars are written next to the output path. Fancy subtitles use .ass; simple subtitles use .srt.

Rendering

Option	Default	Purpose
--codec TEXT	libx264	FFmpeg video codec.
--crf INTEGER	18	Constant rate factor from 0 to 51; lower means higher quality and larger files.
--preset VALUE	medium	FFmpeg encoding preset: ultrafast, superfast, veryfast, faster, fast, medium, slow, slower, or veryslow.
--audio-codec TEXT	aac	FFmpeg audio codec.
--audio-bitrate TEXT	192k	FFmpeg audio bitrate.

Rendering requires both ffmpeg and ffprobe on PATH.

Cache and logging

Option	Default	Purpose
--no-cache	off	Disable cache reads and writes for this run.
--clear-cache	off	Clear cached analysis for each input before processing it.
-v, --verbose	off	Show verbose logging and exception details.
--version	n/a	Print the installed QuickEdit version and exit.
--help	n/a	Print command help and exit.

Analysis caches live in .quickedit_cache/ beside each input video.