render-podcast-skit

Name: render-podcast-skit
Author: gooseworks-ai

1.0kstars193forks

How to install render-podcast-skit?

gooseworks-ai/goose-skills/render-podcast-skit

$npx -y skills add gooseworks-ai/goose-skills --skill render-podcast-skit

Installs into the current project.

Summary

Assemble a two-host fake-podcast skit ad from a config — per-line lipsync clips hard-concatenated in script order, scaled/padded to 1080×1920, WHITE bottom-center captions (up to 5 words per cue, broken on sentence punctuation, word-wrapped to stay in-frame, held at least 0.9s) b

Files · 1

View on GitHub

SKILL.md

1	# render-podcast-skit
2
3	Assemble a two-host fake-podcast skit ad from a config: a skeptic and a believer at an
4	absurd themed podcast desk do a snappy back-and-forth about the product (the set is
5	deliberately unrelated — that is the joke). Each line is its own lipsync clip so the edit can
6	cut on the dialogue beat (~1.8s avg); this capability is the FREE, deterministic assembly
7	that concatenates those clips, renders the WHITE captions, and appends the brand end card.
8
9	`scripts/config.example.json` is the worked example (Ladder run-02 "Laundromat 2am", ~49s
10	1080×1920 9:16, ~22 lines); `scripts/PIPELINE.md` maps every config block to its source step
11	and `scripts/README.md` documents the free assembly.
12
13	## Run
14
15	This is the FREE, deterministic assembly stage — it spends nothing. The paid inputs are
16	separate capabilities: one ElevenLabs with-timestamps VO per line (one voice per host) via
17	`create-vo-elevenlabs`; two photoreal base stills at the themed desk plus ~10 expression variants
18	(mouths NEUTRAL/CLOSED, gpt-image-2 quality=high, not nano-banana) via
19	`create-image-gpt-image-fal`; and one lipsync clip per (still, VO) pair via `create-video-fal`.
20	Given the per-line clips + their VO timestamps + the brand wordmark SVG, `render-podcast-skit`
21	walks the scenes in script order, builds the global caption timeline, renders the WHITE captions,
22	hard-concats the clips, auto-appends the end card, and final-encodes crf28 → the master. Re-cuts
23	reuse the existing VOs / stills / clips and cost $0.
24
25	## Contract (the free assembly)
26
27	- Dialogue-carried, no music bed by default. The per-line VO is the audio; a podcast skit
28	needs no music (an optional low ambience is a taste call, off by default).
29	- One line = one scene = one hard cut, in script order. Hard-concat the per-line clips in
30	order (scale/pad to 1080×1920, re-encode) — no dissolves.
31	- Captions from the VO's OWN char-level timestamps, not Whisper (script-window). Build a
32	global `words.json` by offsetting each line's char-level word timings by the cumulative clip
33	start, group into ≤5-word cues broken on sentence-final punctuation, and render **WHITE
34	`#FFFFFF` bottom-center captions (black outline), word-wrapped to stay in-frame** and held
35	≥0.9s — PIL PNG overlays when the host ffmpeg lacks libass (common), else ASS. (Yellow 3-word
36	karaoke was the old style, rejected in testing.) Whisper on the rendered clips mistimes; the VO
37	timestamps are ground truth.
38	- End card via Playwright/PIL from the real wordmark — never AI-render brand text. The
39	lockup is a deterministic HTML → PNG → 2.5s mp4 from the brand's real wordmark SVG (black bg,
40	brand wordmark, CTA pill, URL), auto-appended after the last line. A diffusion model garbles
41	a wordmark.
42	- FFmpeg composite, deterministic, FREE. Concat the clips, overlay the WHITE caption PNGs (or
43	burn ASS via libass), append the end-card mp4, and **final-encode `-preset slow -crf 28` + aac
44	96k** → a 1080×1920 h264+aac master (~6MB for ~28s; the old `-crf 20` produced ~16MB). No paid
45	calls, no keys.

People who liked this also viewed

6 picks

Type

render-podcast-skit

1.0kstars193forks

View on GitHub

How to install render-podcast-skit?

gooseworks-ai/goose-skills/render-podcast-skit

$npx -y skills add gooseworks-ai/goose-skills --skill render-podcast-skit

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# render-podcast-skit
2
3	Assemble a two-host fake-podcast skit ad from a config: a skeptic and a believer at an
4	absurd themed podcast desk do a snappy back-and-forth about the product (the set is
5	deliberately unrelated — that is the joke). Each line is its own lipsync clip so the edit can
6	cut on the dialogue beat (~1.8s avg); this capability is the FREE, deterministic assembly
7	that concatenates those clips, renders the WHITE captions, and appends the brand end card.
8
9	`scripts/config.example.json` is the worked example (Ladder run-02 "Laundromat 2am", ~49s
10	1080×1920 9:16, ~22 lines); `scripts/PIPELINE.md` maps every config block to its source step
11	and `scripts/README.md` documents the free assembly.
12
13	## Run
14
15	This is the FREE, deterministic assembly stage — it spends nothing. The paid inputs are
16	separate capabilities: one ElevenLabs with-timestamps VO per line (one voice per host) via
17	`create-vo-elevenlabs`; two photoreal base stills at the themed desk plus ~10 expression variants
18	(mouths NEUTRAL/CLOSED, gpt-image-2 quality=high, not nano-banana) via
19	`create-image-gpt-image-fal`; and one lipsync clip per (still, VO) pair via `create-video-fal`.
20	Given the per-line clips + their VO timestamps + the brand wordmark SVG, `render-podcast-skit`
21	walks the scenes in script order, builds the global caption timeline, renders the WHITE captions,
22	hard-concats the clips, auto-appends the end card, and final-encodes crf28 → the master. Re-cuts
23	reuse the existing VOs / stills / clips and cost $0.
24
25	## Contract (the free assembly)
26
27	- Dialogue-carried, no music bed by default. The per-line VO is the audio; a podcast skit
28	needs no music (an optional low ambience is a taste call, off by default).
29	- One line = one scene = one hard cut, in script order. Hard-concat the per-line clips in
30	order (scale/pad to 1080×1920, re-encode) — no dissolves.
31	- Captions from the VO's OWN char-level timestamps, not Whisper (script-window). Build a
32	global `words.json` by offsetting each line's char-level word timings by the cumulative clip
33	start, group into ≤5-word cues broken on sentence-final punctuation, and render **WHITE
34	`#FFFFFF` bottom-center captions (black outline), word-wrapped to stay in-frame** and held
35	≥0.9s — PIL PNG overlays when the host ffmpeg lacks libass (common), else ASS. (Yellow 3-word
36	karaoke was the old style, rejected in testing.) Whisper on the rendered clips mistimes; the VO
37	timestamps are ground truth.
38	- End card via Playwright/PIL from the real wordmark — never AI-render brand text. The
39	lockup is a deterministic HTML → PNG → 2.5s mp4 from the brand's real wordmark SVG (black bg,
40	brand wordmark, CTA pill, URL), auto-appended after the last line. A diffusion model garbles
41	a wordmark.
42	- FFmpeg composite, deterministic, FREE. Concat the clips, overlay the WHITE caption PNGs (or
43	burn ASS via libass), append the end-card mp4, and **final-encode `-preset slow -crf 28` + aac
44	96k** → a 1080×1920 h264+aac master (~6MB for ~28s; the old `-crf 20` produced ~16MB). No paid
45	calls, no keys.

People who liked this also viewed

6 picks

Type