Predictable defaults

Profile	Behavior
fast (default)	Container metadata, streams, chapters, and embedded subtitles. No scenes, frames, or speech recognition
quality	Adds ffmpeg scene detection. It does not extract frames or run speech recognition
Transcript	Off. include_transcript on the shipped server returns a warning and an empty transcript
OCR	video_evidence only, not part of quality
Frames	A named video_evidence follow-up. include_keyframes on the shipped server returns a warning and no frames

Missing ffprobe or embedded subtitles is reported as a gap. No cloud video API or frame-by-frame vision model is required.