gooseworks-ai/goose-skills / skills/video/render-absurdist-explainer
render-absurdist-explainer
VerifiedAssemble an absurdist animated-explainer video ad (~38s, 9:16) from per-scene i2v clips + their measured VO windows — retime each clip to its VO, re-encode every segment to identical 30fps/libx264/yuv420p so the concat demuxer never drops frames, concat, build a REAL-product PIL end card (never AI) held static, mix VO (loudnorm I=-14) under music (loudnorm I=-26, volume 0.62, amix normalize=0), and burn libass captions last. FREE deterministic assembly (bash-free, Python + ffmpeg + PIL); the recipe supplies the clips, VO, music, product photo, palette, and caption table and gates the paid keyframe/clip/VO/music calls to their own capabilities. Use for the absurdist-explainer format.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 9.5 KB |
| scripts/build_endcard.py | script | 7.3 KB |
| scripts/compose.py | script | 16.5 KB |
| scripts/config.example.json | config | 4.5 KB |
| scripts/make_captions.py | script | 3.4 KB |
| skill.meta.json | config | 236 B |
| tests/smoke-test.md | markdown | 4.2 KB |
tests/smoke-test.md
View on GitHub# Smoke test — render-absurdist-explainer
Verifies the free PIL + ffmpeg assembly end-to-end. No paid calls. Needs: Python 3 with
Pillow, ffmpeg/ffprobe, and a sans-serif TrueType font (DejaVu on most Linux, Arial on
macOS, Arial or Segoe UI on Windows). You supply per-scene clips + VO mp3s + a music bed + a real
product photo (any 1080×1920 mp4s and a ≥1000px product jpg work for a smoke run).
## Setup
```bash
cd scripts
python3 -m pip install pillow # if not present
mkdir -p /tmp/absurdist-smoke
# Copy config.example.json -> config.json and point every path at real files:
# - scenes[].clip : your per-scene i2v mp4s (1080x1920)
# - scenes[].vo : the per-scene VO mp3s (target_sec = ffprobe of each)
# - end_card.product_image : a real retail product photo (>=1000px)
# - music_bed : an instrumental bed (or set to null to run VO-only)
# - end_card.vo : the end card's spoken line (or remove the key for a silent card)
# - captions + end_card copy + brand_palette : your brand's (the example's are a
# worked example, not defaults)
cp config.example.json /tmp/absurdist-smoke/config.json
```
## Run
```bash
# 1) real-product end card (PIL) — must run BEFORE compose
python3 build_endcard.py --config /tmp/absurdist-smoke/config.json \
--out /tmp/absurdist-smoke/endcard.png
# (then set end_card.image in config.json to /tmp/absurdist-smoke/endcard.png)
# 2) per-scene captions (libass) — must run BEFORE compose
python3 make_captions.py --config /tmp/absurdist-smoke/config.json \
--out /tmp/absurdist-smoke/captions.ass
# (then set captions_ass in config.json to /tmp/absurdist-smoke/captions.ass)
# 3) assemble the master
python3 compose.py --config /tmp/absurdist-smoke/config.json \
--work-dir /tmp/absurdist-smoke \
--out /tmp/absurdist-smoke/master.mp4
```
## Expect
- `build_endcard.py` writes `endcard.png` (1080×1920): the REAL product photo centred over
the brand palette, with a typeset wordmark + claim rows + accent CTA pill. NOT an AI
bottle; NO smeared/AI-rendered text.
- `make_captions.py` writes `captions.ass` — one cue per scene with a caption, the end card
suppressed, `start = scene_start + 0.08s`.
- `compose.py` prints per-scene retime lines, the total runtime, a `loudness:` line, and
a final `WROTE ... (expected ~Xs, delta ...)`. `master.mp4` is 1080x1920, 30fps; its
duration is within 0.1s of `sum(scenes[].target_sec)` + the end-card dwell.
- The `loudness:` line reads between -14.5 and -13.5 LUFS, with a true peak <= -1.5 dBFS.
- With `end_card.vo` set, the line is heard over the end card. If the line + 0.5s is
longer than `dwell_sec`, compose prints the stretched dwell.
- The caption burn works with an absolute `captions_ass` path, on Windows too.
- Run the `watch` skill on `master.mp4`: every character's look holds across scenes,
the single narrator voice (whoever the recipe's `choices.narrator` picked) carries the
whole spot, no AI brand text
leaked into a cartoon background, captions don't collide with on-screen text, and the end
card is the real product with legible copy.
## Fail signals
- Concat drops frames / audio desyncs → a segment wasn't re-encoded to 30fps (all segments
MUST be `libx264 -r 30` before the concat demuxer). compose.py always re-encodes, so this
means a source clip fed the wrong stream — check the ffprobe output.
- compose.py warns that the master is outside the loudness target -> the master pass
could not converge. Check the VO and music files are not silent or clipped.
- End card shows a cartoon/AI bottle or smeared text → `end_card.product_image` points at
an AI render, or a font failed to load (build_endcard tries DejaVu, Arial, Segoe UI,
then Pillow's built-in font and prints a warning - install DejaVu or Arial).
- End-card text is a different colour from the palette -> the brand colour was too close
to the background. build_endcard printed a contrast warning; set `end_card.background`
or change the palette.
- Caption flashes a frame before a cut → the +0.08s offset was removed from make_captions.
- Duration far off the summed windows → a `target_sec` wasn't the measured VO duration
(ffprobe each VO mp3; don't use planned word counts).
```
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| receiving-code-reviewobra/superpowersUse when receiving code review feedback, before implementing suggestions, especially if feedback seems unclear or technically questionable - requires technical rigor and verification, not performative agreement or blind implementation | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| requesting-code-reviewobra/superpowersUse when completing tasks, implementing major features, or before merging to verify work meets requirements | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| using-git-worktreesobra/superpowersUse when starting feature work that needs isolation from current workspace or before executing implementation plans - ensures an isolated workspace exists via native tools or git worktree fallback | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| verification-before-completionobra/superpowersUse when about to claim work is complete, fixed, or passing, before committing or creating PRs - requires running verification commands and confirming output before making any success claims; evidence before assertions always | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |