gooseworks-ai/goose-skills / skills/video/compose-creator-layer
compose-creator-layer
VerifiedComposite a generated creator track into a product layer, per beat — in the creator's zone on split beats (either side of a seam), full frame on creator beats, hidden on product beats — with the creator's voice as the audio (loudness -14 LUFS) and an optional ducked music bed, or a music-only or silent master for voiceless formats. Free, local ffmpeg. Pairs with footage-cutlist (the layer) and create-creator-takes-h3 (the creator). Use for split-screen (with an optional zoom and brand-colour seam bar), screen-insert and any creator-over-product format.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 3.2 KB |
| scripts/compose.py | script | 7.2 KB |
| skill.meta.json | config | 206 B |
| tests/smoke-test.md | markdown | 383 B |
| tests/test_compose_split.py | script | 2.9 KB |
scripts/compose.py
View on GitHub#!/usr/bin/env python3
"""Drop the creator into the product layer: the creator's zone on split beats, the whole
frame on creator beats, hidden on product beats. The creator's voice is the audio. Free.
compose.py --layer layer.mp4 --beats cutlist.aligned.json --creator creator.mp4 \
--out reel.mp4 [--music bed.mp3 --music-db -20] [--head 0.30] \
[--zoom 1.15] [--divider '#3366FF' --divider-px 4]
--layer is footage-cutlist cut.py output, rendered from the SAME (aligned) cut list as
--beats, so the two timelines agree. --creator is create-creator-takes-h3 join_takes.py
output (one continuous creator track, reel time from 0).
The creator is scaled to COVER each area and cropped, with the head kept near the top
third (--head is where the crop sits vertically: 0 = keep the top, 0.5 = centre). The crop
is computed from the take's real size; a hard-coded input height once cropped the top half
of an upscaled take and blew it up.
Split screen (the split-screen creator look): --zoom pushes in on the creator inside their
zone so a webcam-distance take reads head-to-shoulders (1.15 to 1.2; tune by eye, it is a free
re-run), and --divider draws a brand-colour bar across the seam on split beats only.
NO GRADE. Matching the take's saturation or warmth to a number was tried twice on the
reference build and both looked wrong; the take goes in as generated.
Audio (--audio):
creator (default) the creator's own voice, loudness-normalised to -14 LUFS; an optional
--music bed is mixed under it and ducked (sidechain) so the voice always leads.
music the --music bed alone (a voiceless format, e.g. a screen walkthrough for a
paid placement), normalised to -14 LUFS with a fade out.
none a SILENT master (organic posts where the track is added in-platform at upload,
which is licensed for organic use; a paid ad cannot use that library).
The creator track may be shorter than the reel (a 2.3s reaction hook): it only shows during
its own `creator` beats.
"""
import argparse
import json
import pathlib
import re
import subprocess
def zone(spec):
W, H = spec["size"]
seam = spec.get("seam") or H // 2
if spec.get("creator_side", "bottom") == "bottom":
return seam, H - seam
return 0, seam
def even(n):
return max(2, int(round(n / 2.0)) * 2)
def cover(label, out, w, h, head, zoom=1.0):
"""Scale to cover a w x h area (times zoom, to push in) and crop it, the crop's top at head."""
sw, sh = even(w * zoom), even(h * zoom)
return ("%s scale=%d:%d:force_original_aspect_ratio=increase:flags=lanczos,"
"crop=%d:%d:(iw-%d)/2:(ih-%d)*%.3f,setsar=1%s" % (label, sw, sh, w, h, w, h, head, out))
def divider(seam, px, color, wins):
"""A colour bar centred on the seam, drawn only while the split beats show."""
return "drawbox=x=0:y=%d:w=iw:h=%d:color=%s@1:t=fill:enable='%s'" % (
seam - px // 2, px, color, enable(wins))
def windows(beats, state):
return [(b["start"], b["end"]) for b in beats if b.get("state", "split") == state]
def enable(wins):
return "+".join("between(t,%.3f,%.3f)" % (s, e - 0.001) for s, e in wins) or "0"
def main():
ap = argparse.ArgumentParser()
ap.add_argument("--layer", required=True)
ap.add_argument("--beats", required=True)
ap.add_argument("--creator", required=True)
ap.add_argument("--out", required=True)
ap.add_argument("--music")
ap.add_argument("--music-db", type=float, default=-20.0)
ap.add_argument("--head", type=float, default=0.30)
ap.add_argument("--audio", choices=["creator", "music", "none"], default="creator")
ap.add_argument("--zoom", type=float, default=1.0,
help="push in on the creator inside the split zone (1.0 = cover only)")
ap.add_argument("--divider", help="brand colour of the seam bar on split beats, #RRGGBB")
ap.add_argument("--divider-px", type=int, default=4, help="seam bar height in px (even)")
a = ap.parse_args()
if not 1.0 <= a.zoom <= 2.0:
raise SystemExit("--zoom must be between 1.0 and 2.0")
if a.divider is not None and not re.fullmatch(r"#[0-9A-Fa-f]{6}", a.divider):
raise SystemExit("--divider must be a #RRGGBB colour")
if a.divider_px < 2 or a.divider_px % 2:
raise SystemExit("--divider-px must be an even number of pixels, 2 or more")
spec = json.loads(pathlib.Path(a.beats).read_text(encoding="utf-8"))
lj = pathlib.Path(a.layer).with_suffix(".json")
if lj.exists() and json.loads(lj.read_text()).get("draft"):
raise SystemExit("%s is a --draft layer; re-run cut.py without --draft" % a.layer)
W, H = spec["size"]
fps = int(spec.get("fps", 30))
total = spec["beats"][-1]["end"]
zy, zh = zone(spec)
split_w, full_w = windows(spec["beats"], "split"), windows(spec["beats"], "creator")
fc = ["[1:v]trim=0:%.3f,setpts=PTS-STARTPTS,fps=%d,split=2[ca][cb]" % (total, fps),
cover("[ca]", "[cz]", W, zh, a.head, a.zoom),
cover("[cb]", "[cf]", W, H, a.head),
"[0:v]trim=0:%.3f,setpts=PTS-STARTPTS,setsar=1[bg]" % total,
"[bg][cz]overlay=0:%d:enable='%s'[v1]" % (zy, enable(split_w)),
"[v1][cf]overlay=0:0:enable='%s'%s,format=yuv420p[v]" % (
enable(full_w),
"," + divider(spec.get("seam") or H // 2, a.divider_px, a.divider, split_w) if a.divider else ""),
]
inputs = ["-i", a.layer, "-i", a.creator]
amap = ["-map", "[a]"]
if a.audio == "creator":
fc.append("[1:a]atrim=0:%.3f,asetpts=PTS-STARTPTS,aresample=48000[vo]" % total)
if a.music:
inputs += ["-stream_loop", "-1", "-i", a.music]
fc += ["[vo]asplit=2[vo1][vo2]",
"[2:a]atrim=0:%.3f,asetpts=PTS-STARTPTS,aresample=48000,volume=%.1fdB,"
"afade=t=out:st=%.3f:d=0.8[m]" % (total, a.music_db, max(0, total - 0.8)),
"[m][vo2]sidechaincompress=threshold=0.03:ratio=6:attack=20:release=300[md]",
"[vo1][md]amix=inputs=2:duration=first:normalize=0,loudnorm=I=-14:TP=-1.5:LRA=11[a]"]
else:
fc.append("[vo]loudnorm=I=-14:TP=-1.5:LRA=11[a]")
elif a.audio == "music":
if not a.music:
raise SystemExit("--audio music needs --music")
inputs += ["-stream_loop", "-1", "-i", a.music]
fc.append("[2:a]atrim=0:%.3f,asetpts=PTS-STARTPTS,aresample=48000,afade=t=out:st=%.3f:d=0.8,"
"loudnorm=I=-14:TP=-1.5:LRA=11[a]" % (total, max(0, total - 0.8)))
else:
amap = ["-an"]
subprocess.run(["ffmpeg", "-v", "error", "-y", *inputs, "-filter_complex", ";".join(fc),
"-map", "[v]", *amap, "-t", "%.3f" % total, "-r", str(fps),
"-c:v", "libx264", "-crf", "12", "-preset", "medium", "-c:a", "aac", "-b:a", "192k",
"-ar", "48000", "-movflags", "+faststart", a.out], check=True)
d = float(subprocess.run(["ffprobe", "-v", "error", "-show_entries", "format=duration", "-of", "csv=p=0",
a.out], capture_output=True, text=True).stdout)
print("[compose] %.2fs %dx%d creator zone y=%d h=%d split %d / full %d / product %d beats -> %s"
% (d, W, H, zy, zh, len(split_w), len(full_w),
len(spec["beats"]) - len(split_w) - len(full_w), a.out))
if __name__ == "__main__":
main()
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack." | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |