video-reframe · Smart Aspect-Ratio Conversion

Converts video to a target aspect ratio — 9:16, 16:9, 1:1, and more — with three strategies: blurred-background fill, focus crop, and face-centered crop.

Horizontal-to-vertical defaults to blurred fill (nothing lost from the frame); talking-head videos can use face-aware cropping; when smart mode finds no face it degrades to center crop and says so in the report.

Example invocation: "Convert this landscape video to 9:16 vertical."

Full brief

Positioning

video-reframe is the atomic tool for aspect-ratio conversion: video in, target aspect ratio out, three strategies to choose from — all through a deterministic script, never hand-built filters.

Core capabilities

• blur (default): original frame centered with its own blurred, upscaled copy as background — no black bars, the safest horizontal-to-vertical conversion, nothing lost.
• crop: crop to the target ratio around a focal point — no black bars but edges trimmed; --focus-x pins the focus when the subject is off-center (0 = left, 1 = right).
• smart: cv2 face detection, cropping around the median face position; degrades to center crop when no face is found, noted in the report.
• Target ratios: 9:16 / 16:9 / 1:1 / 4:5 / 3:4 / 4:3 / 21:9; --size forces the final resolution.
• Audio track preserved automatically (copy, no audio re-encode); output resolution forced to even alignment (H.264 requirement) — no manual handling.

Workflow

1. Confirm the video file, target ratio, and mode (default blur when unsure)
2. Run reframe.py reframe — the script computes crop geometry, focus bounds, and even alignment
3. Receive the converted video plus a report (source→target dimensions, strategy used, face focus for smart mode)

Inputs & outputs

Output: converted video (*-<ratio>.mp4) under outputs/<topic>/ plus the conversion report.

Boundaries with adjacent skills

• video-editing aspect: simple padding/cropping, black bars, center crop — lightweight and general; reach for this skill when a strategy-driven conversion is needed.
• clipify: per-shot dynamic face-tracking pan, used inside the clipping pipeline; this skill reframes whole segments.
• video-reframe (this): three-strategy intelligent conversion — the safest pick for horizontal-to-vertical.

Fit

• Adapting landscape video for Douyin/Channels vertical publishing
• Face-centered vertical conversion of talking-head videos
• Square, 4:5, and other multi-platform size adaptations

Before you start

• No API key; the SKILL.md lists no extra configuration requirements.
• Mode rule of thumb: blur first for horizontal-to-vertical (nothing lost); crop when the subject is clear and should fill the screen; smart for talking-head content.
• With a Profile, the default target ratio follows the primary platform (Douyin/Xiaohongshu 9:16, Bilibili 16:9, Moments/IG 1:1); without one, default blur and ask for the target platform.

video-reframe is part of the Aiglade Skill library. Invoke it from the Aiglade chat box in plain language.

← All 155 skills