Skip to content
cccskills
FreePublished on GitHub

render-imessage-chat

Render a configurable iMessage-style conversation inside a framed phone, followed by a brand end card. Text, typing, scrolling, and original send and receive sounds share one timeline. Local Playwright and ffmpeg assemble the video without paid APIs for rendering.

OriginalShow the original description

Render a configurable iMessage conversation inside a properly framed phone, then a brand end card. Uses the original send/receive sounds and a shared frame timeline for text, typing, scrolling and sound. Free local Playwright + ffmpeg assembly; optional image/music generation belongs to separate gated capabilities.

gooseworks-ai/goose-skills1,272 stars(GitHub 1,272 + cccskills 0)

Run score Not tested

See how to install

Good for

  • Making a texting-story ad
  • Rendering a branded phone chat video
  • Showing a conversation with custom names and time
  • Creating a vertical ad with an end card

Files included (24)

  • SKILL.md12.8 KB
  • assets/sfx/imessage-receive.mp311.5 KB
  • assets/sfx/imessage-send.mp34.2 KB
  • references/imessage-reference.md5.2 KB
  • scripts/chat-driver.js6.3 KB
  • scripts/check-render.py6.1 KB
  • scripts/config.example.json1.7 KB
  • scripts/end-card.template.html3.5 KB
  • scripts/mockup/generate.js15.9 KB
  • scripts/mockup/templates/chat.css18.1 KB
  • scripts/mockup/templates/icons.js3.3 KB
  • scripts/package-lock.json1.7 KB
  • scripts/package.json322 B
  • scripts/record-chat.js25.0 KB
  • scripts/render-end-card.js14.3 KB
  • scripts/render.sh992 B
  • scripts/sfx-embedded.json21.7 KB
  • scripts/stitch.sh8.9 KB
  • skill.meta.json230 B
  • tests/fixtures/long-group-thread.json3.3 KB
  • tests/smoke-test.md1.5 KB
  • tests/test_chat.js12.3 KB
  • tests/test_safe_area.js10.1 KB
  • tests/test_stitch.py12.9 KB

SKILL.md (original)

Review the instructions the agent will receive before you install.

Superseded: the video kit now does this with the phone-chat part, version 1.2.0, in the parts folder of this repository. Every phone-chat style is a style file that draws the same screen frame by frame from the plan's scenes, with the original sounds, in the kit's browser. This atom stays, unchanged in behaviour, for skills outside the kit until they move; its scripts still run.

Human version

Make a texting-story ad with the chosen contact names and phone time, a proportional phone, an inset Dynamic Island and the original iMessage sounds. The brand, story and background can change. Images are optional and can appear anywhere in the conversation; product links and music are optional too.

This rebuild preserves the shell and sounds from the approved Dermwise Goa build. It fixes missing identity binding, the island touching the screen edge, light-mode header colors, early/missing group names and capture timing drift. Every movie frame and sound cue uses the same timeline; browser startup cannot trim the beginning or ending. No paid API is needed to render or repair the UI.

On a 9:16 canvas the phone now sits clear of the TikTok/Reels controls by default: the newest message, including the punchline, always stays above the bottom caption band and left of the button rail (safe_area).


Agent version

Read the reference and authoring rules. The calling recipe supplies NEW names, copy, brand facts and real assets. scripts/config.example.json is a fictional, runnable example, never defaults.

Choices and bindings

  • Relationship, contact names and group title → thread.participants, thread.title.
  • Displayed phone time → thread.clock, preserving the user's chosen text.
  • Story, tone and language → thread.messages. Casual spelling and emojis are allowed.
  • Chat images → zero or more attachment entries at their authored positions in thread.messages; either participant may send them. Never reorder by type.
  • Theme → theme: "dark" | "light".
  • Background → optional local background_image; neutral when absent.
  • Hardware → dynamic_island: true | false; true floats inside the screen.
  • Music → optional existing bed passed to render.sh --music; no bed means SFX only.

Exactly one participant is self:true. A DM has two participants; its header reads the other participant's name and derives the first initial unless supplied. A group requires a title and named contacts. Missing names fail before capture. There is no demo-name fallback. Changing the config changes the visible name. When a name changes, update any derived initials too. Keep explicit user-supplied initials only when they still match the requested identity.

Run

Requires Node 18+, Python 3, ffmpeg with libx264, ffprobe and Playwright Chromium. Install dependencies in the fetched scripts folder, then launch/close that script's own Chromium before any optional paid image/music call. Preserve its cwd, NODE_PATH and PLAYWRIGHT_BROWSERS_PATH. gooseworks doctor --renderer-script "/absolute/path/scripts/record-chat.js" can check that runtime. If unavailable, use a bounded free createRequire(actualScript) launch/close probe (15-second launch timeout, 20-second whole-process limit). Cache presence alone is not proof.

cd scripts
npm ci
# Install Chromium only if the free launch probe says it is missing:
# npx playwright install chromium
node record-chat.js --config /absolute/path/config.json --out-dir /absolute/path/working/preview --preview-only
bash render.sh --config /absolute/path/config.json --out /absolute/path/finals/master-final.mp4
# Optional: append --music /absolute/path/bed.mp3

Preview produces chat.html, chat-preview.png and master-chat.safe-area.json; the HTML exposes window.__renderAt(seconds) and window.__safeAreaReport() (canvas-pixel boxes of the newest row and any data-safe-keep sheet or dialog) for frame inspection. Full render keeps those, master-chat.mp4, .timeline.json, .sfx.json, the end-card HTML/PNG/MP4 and the finished master. The recorder checks the safe area on every output frame and fails on a violation. check-render.py verifies dimensions, audio stream, frame count, complete ending and the safe-area report; python3 scripts/check-render.py --safe-area <work>/master-chat.safe-area.json checks a preview alone. Review the ACTUAL master after every repair; these technical checks do not establish creative acceptance.

Individual record-chat.js, render-end-card.js and stitch.sh commands remain available. render.sh produces the 1080×1920 recipe master. The chat recorder also supports even preview dimensions; the phone must fit with a margin.

Config contract

  • Inline thread or thread_path. Relative files resolve against config.json.
  • Unique message IDs and valid from participants. Types: text, typing, timestamp, attachment, tapback. Reactions target an earlier message ID and carry an emoji. Typing immediately precedes a received text/attachment from the same person. Self messages type in the composer, including complete emoji graphemes.
  • Short messages read best. Longer words wrap; real overflow fails preflight.
  • Optional attachment: src local image/data URI, presentation:"photo" for a photo or "rich-link" for image + flush meta card + title/domain/chevron. Optional dwell_sec overrides its default 3.6-second reading hold. Text-only chats need no images. One or several attachments can come first, between any messages or last; preserve the user's placement and sender. Never require an opening photo or a product image at a fixed beat.
  • Optional thread.clock sets the displayed status-bar time, such as 10:24 or 18:07. Bind the chosen time; do not replace it with a demo time. Only when absent does the shell use its neutral 9:41 fallback. In-chat timestamp labels are separate message inputs, not a required fixed timestamp.
  • End card: approved image_path, or real logo_svg_path, logo_image_path, inline logo_svg or wordmark_text, brand colors, CTA and optional benefits. stars defaults to 0. Ratings require approved proof_text. An artwork path replaces the complete template; check its copy and CTA first.
  • Default outer canvas 1080×1920; zoom fits the 393×852 phone proportionally. Excess zoom fails rather than cropping the phone. timing can override the named pacing fields in record-chat.js; ending hold must be at least 0.5 seconds.
  • safe_area keeps the conversation clear of the platform controls. Omitted: on for 9:16 canvases, off for other shapes. true uses the review-finished-ad bands (top 220, bottom 400, right 140, left 0 px at 1080×1920, scaled to the canvas). An object such as {"bottom":480} overrides single bands in output pixels; omitted keys keep the defaults. false restores the old centred full-height phone exactly.
  • With safe_area on, the phone is the largest proportional size whose conversation viewport sits inside the zone (an 8 px inset) while the phone stays in the middle of the canvas. At 1080×1920 that is zoom about 1.648 with 258 px above and below the phone. phone_fit: "largest" gives the biggest safe phone instead (zoom about 1.97), which has to sit 16 px from the top with a wide gap under it. Every row is clipped to that viewport, so the newest row is safe on every frame. The composer, home bar and group avatars may sit in the bands; typed text reappears as the newest row.
  • An explicit zoom is a ceiling while safe_area is on: kept when safe, otherwise lowered to the safe maximum with a log line (the recipe seed zoom: 2.1 becomes about 1.648). Set safe_area:false to keep it exactly.
  • The framed phone draws the home indicator, with the 34 px strip iOS keeps under the composer for it.
  • timing.tail_hold (default 2 s) is how long the finished thread holds before the end card. With the gap that follows the last message, the last line is on screen about 2.7 s; at 1 s a nine-word punchline had 1.7 s.

Original sound contract

The send/receive MP3s are byte-identical to both archived Dermwise and Wonderloom builds. Keep them. No substitute ringtone, notification-cascade sound or generated pop. One cue per real text/attachment; none for typing or composer keystrokes. Picture reveals and cues share fixed output-frame times; the audible onset follows the first visible reveal frame. The mixer strips leading silence and limits peaks.

Full checkouts use assets/sfx. Text-only catalog packages use the hash-checked scripts/sfx-embedded.json fallback. Keep that file byte for byte. --sfx-dir can override the source explicitly. Missing, silent, corrupt or LFS-pointer audio stops the render. After an intentional MP3 replacement, a builder regenerates the embedded copy in the goose-skills repo with python3 tests/test_stitch.py --write-embedded.

Verification and failures

The tests live in the goose-skills repo, not in installed copies. There, run node --test tests/test_chat.js, node --test tests/test_safe_area.js and python3 -m pytest tests/test_stitch.py. Run one test file at a time; each test opens and closes one Chromium. Browser tests cover changing names/time/background, text-only chats, attachment placement, blank-name rejection, inset hardware, dark/light chrome, group labels after typing, Unicode composer text and long threads. test_safe_area.js walks every frame of tests/fixtures/long-group-thread.json (16-message group thread, wrapped punchline) and asserts the newest row stays above y 1520 and left of x 940, that a bottom sheet in the band fails, and that safe_area:false keeps the old layout. CI runs both browser files. Audio tests cover fetched-package delivery and limited overlapping cues.

Fix the configuration error and rerender locally. UI defects never justify paid generation. Watch the final for the selected name, readable bubbles, smooth scroll, correct sender labels and sounds, complete last message and correct brand end card. Use review-finished-ad for brand/copy review when called by the recipe.

Critical knowledge

The current renderer combines the fixed-frame repair with the lessons from the live-capture audit. Read [[references::references/imessage-reference.md]] before authoring.

  1. Browser startup and CPU load must never change movie time. Fixed output frames replace capture-clock guesses, sync curtains and picture-snapping retries.
  2. Measure each sound's audible onset. The original send file includes lead-in; trim silence before placing it on the visible reveal frame.
  3. Keep the original receive chime. Shorten it only when another message follows quickly, so its second note cannot mask that next message.
  4. Check the text Range against the bubble bounds. Bubble tails intentionally extend beyond the box.
  5. Use Apple emoji assets for recordings on hosts whose native emoji differ. Cache and inline them before capture; keep complete Unicode graphemes while typing.
  6. Scratch directory templates must work on macOS and GNU systems.
  7. Start short conversations under the header. Keep the input fixed at the bottom and scroll only the conversation.
  8. Picture and sounds share output-frame time. Reactions use that same timeline and must target an earlier real message.
  9. Do not zoom into a link as if a camera were moving across the phone screen.
  10. Editorial endings use the brand's own fonts, headlines, benefits, URL and footnote. Approved complete artwork can replace the template.
  11. Show Delivered only beneath the newest sent text or attachment.
  12. Typed text must equal sent text. The deterministic composer completes before sending and wraps long lines.
  13. Download and inline requested end-card fonts before capture. A missing font fails the render instead of silently changing the brand.
  14. Read approved brand colours from the brand kit or site styling. Preserve the selected background and contact names.
  15. Keep the quieter audit mix with the existing peak limiter. Unsupported ratings remain absent unless approved proof is supplied.
  16. Inspect the actual encoded ending and sound alignment. Frame counts and a passing stream probe do not establish creative acceptance.
  17. Keep the newest message out of the platform controls. A full-height phone put the punchline under the TikTok/Reels caption band. Fit the conversation viewport, not the whole phone, into the safe zone: the phone stays large and native. A future skin with bottom sheets must extend PHONE.keep to the screen bottom and mark sheets data-safe-keep.

Reviews

No reviews yet. Tell others what you thought.

Comments

No login needed. Share how it worked for you, ask a question, or note where it did not run.

No comments yet. Be the first to write one.

Comments are published right away. Do not include personal information such as email addresses or phone numbers. Up to 2 URLs.

0 / 1,000 characters

This form is protected by reCAPTCHA, and the Google Privacy Policy and Terms of Service apply.

More skills from this repository

<slug>

FreeSummary

The supplied material contains only placeholder text for the skill’s purpose, inputs, steps, and completion criteria. It does not provide enough information to describe a task or result reliably.

  • No use case can be identified from the supplied material
  • Purpose is unspecified
  • Inputs and expected results are unspecified
gooseworks-ai/goose-skills1.3KUpdated Oct 27, 2026

ad-angle-miner

FreeSummary

Research supported advertising angles using buyer context, current product facts, customer language, and observed ads. Turn the evidence into static ad, video, or copy ideas, matching video concepts to available recipes. Scores rank ideas for testing; they do not predict results.

  • Finding evidence-backed angles for a product campaign
  • Developing static ad or video concepts from customer research
  • Writing headline and primary-text options for ads
gooseworks-ai/goose-skills1.3KUpdated Oct 27, 2026

Analyze campaign data from Google, Meta, and LinkedIn to find strong and wasteful spend. The skill checks funnel performance and test results, then recommends which campaigns to pause, scale, or test and how to shift budget across channels.

  • Finding campaigns and ads that waste budget
  • Comparing performance across Google, Meta, and LinkedIn
  • Planning specific budget shifts and follow-up tests
gooseworks-ai/goose-skills1.3KUpdated Oct 27, 2026

Compare ad spend and signups with downstream outcomes to calculate the cost of each qualified lead. The skill checks tracking coverage, accounts for how long leads need to mature, and groups creatives into actionable categories with confidence notes.

  • Compare true qualified-lead costs across creatives
  • Find low-CPA ads that bring poor-quality leads
  • Check whether signups can be linked to downstream outcomes
gooseworks-ai/goose-skills1.3KUpdated Oct 27, 2026

Prepare the intake for a Meta ad campaign by reviewing available brand information first, stating reasonable assumptions, and asking only questions whose answers could affect budget, claims, or legal exposure. It records stable brand facts, a campaign brief, and what was asked versus assumed.

  • Prepare a Meta ad campaign when brand facts are missing
  • Clarify budget and timing before campaign planning
  • Record approved claims and audience restrictions
gooseworks-ai/goose-skills1.3KUpdated Oct 27, 2026

ad-policy-check

FreeSummary

Review ad copy, image text, visuals, and people against platform and brand category rules. It suggests honest fixes and needed disclosures, then flags issues requiring human or legal review, with a separate result for each ad and platform.

  • Checking ad copy and visuals before launch
  • Reviewing creator ads for required disclosures
  • Fixing a platform disapproval and recording the issue
gooseworks-ai/goose-skills1.3KUpdated Oct 27, 2026

See all skills by gooseworks-ai

Report a problem with this skill