An Indonesian YouTuber with 1.41M subscribers open-sourced a motion skill on the side: how Bang Motion got 1,094 stars
What it is A skill you install into Claude Code, Codex, Gemini CLI, or Cursor that lets AI use code to make intros, promo videos, and explainers, outputting a single HTML file that plays on double-click.
Why it took off The author is an Indonesian AI tutorial creator with 1.41M subscribers. It debuted in September with one of his videos that got 100k views, riding the "AI motion design" wave kicked off by HyperFrames and Claude Motion. In early November, a batch of GitHub repost accounts pushed it again.
CHow you could get in Don't build the 41th generic animation skill. Creators who study this follow its loop of "each video ships with one of your own open-source libraries." If you do it, pick a narrow niche (vertical product-promo clips, Chinese subtitle and voiceover alignment) and ship within one to two weeks of the next big-company release.
- Data (updated daily) · Repo created 2026-09-05 · 1,502 stars · 7 days +621
GitHub · stars Data 2026-10-09 → 2026-11-08 (31 days, 31 observations). Points before 10-28 come from deeply / archive backfill, not every day has one
Bang Motion is a skill that makes coding agents like Claude Code and Codex produce motion that looks like video rather than slides. It got 1,094 stars in 62 days and gained 332 more in the last week. Most of those stars didn't come from the code. The author runs an AI tutorial channel with 1.41M subscribers, and the launch rode on one of his own promo videos. The early-November wave came from a group of "great GitHub library" repost accounts. The space is already crowded, so anyone considering it should first ask whether they have a channel of their own.
Sep 19 Indonesian video
In Sep 2026 19, the YouTube channel Bang Tutorial posted an Indonesian-language video titled "Bikin Motion Graphic Pakai ChatGPT (GPT-6 Astra + Higgsfield MCP)", meaning "Making motion graphics with ChatGPT." The first line of the description is a Higgsfield promo link. The next paragraph reads: "In this video I introduce Bang Motion, a free open-source agent skill I made myself, that makes AI-generated motion actually move like video instead of a PowerPoint slideshow."
By Nov 6, the video had 103k views and 1,913 likes (YouTube view count, 11-06). The channel itself has 1.41M subscribers, 589 videos, and 189M total views, and lists Indonesia as its country. Five days later, on Sep 24, a website that auto-generates tech news wrote that it had "19 days, 493 stars, 72 forks."
Three days before that, AutoClaw, a channel under Zhipu, also posted an English video, "Made a 4K promo video with the newly open-sourced Bang Motion," which got only 140 views. Whether the two were collaborating is not visible from the outside.
What it actually does
In short: you tell the agent "make a 30-second promo for my product," and it writes an index.html following Bang Motion's rules. Double-click it and it autoplays and loops. Add ?debug=1 and you get a scrubber bar, and you can also export frame by frame as MP4 to sync with voiceover.
The opening paragraph of the README is its entire pitch:
Ask an AI agent for "a video" and you usually get slides.
—When you ask AI for a video, you usually get a slideshow: one idea per page, one title, one image, with fades between pages. It writes down the handful of things motion designers insist on (one continuous world, a camera that really moves, a subject that runs through the whole piece, text that lives inside the scene) as rules the agent can check in its own code.
A few real differences:
- Browser only. GSAP plus CSS. No Remotion, no Node. The output is a single HTML file you can take anywhere.
- Brief first, then code. The agent first gives 3 candidate concepts, one recommendation, and a palette where every color is sourced. You approve before it starts building. After writing, it screenshots key frames and checks them against the rules.
- Menu-style: 10 styles × 19 intro structures × 6 explainer styles.
- It works across four agents (Claude Code, Codex, Gemini CLI, Cursor). It also documents generating assets and voiceover through the Higgsfield MCP, and even building cartoon explainers directly in After Effects.
The README's hero image is a 12-panel style grid, one finished frame per cell. The public repo has only 3 commits, all in Sep 15, and no code has been pushed since.
How the stars grew
| Timeline | What happened | star |
|---|---|---|
| 9-05 | Repo created | 0 |
| 9-19 | Author posted on his own channel, 103k views | — |
| 9-24 | News site coverage | 493 |
| 10-17 ~ 10-31 | Nobody pushed it. Gained 7 to 16 per day | 637 → 778 |
| 10-31 | gittrend.io posted short videos on TikTok and Instagram (992 TikTok views) | 778 |
| 11-01 | Instagram @synthetix_motion posted "4 steps to make animated explainers with Claude Code." Step one is installing Bang Motion. | +103 → 881 |
| 11-02 | Prinon Turio posted introductions on YouTube and TikTok (1,043 YouTube views) | +78 → 959 |
| 11-03 ~ 11-06 | GitFinder (296 views), Brazil's Fabiano Carvalho with a Portuguese intro (668 views), Trending Open Source Projects (857 views) | → 1,094 |
Data: GitHub daily star counts (starting from 10-17) and per-platform view counts (11-06).
The two waves of sources are clear.
Wave one was the author himself. A channel with 1.41M subscribers. One video with 100k views put the repo link in its description, and it went from zero to 19 stars in 493 days. Similar libraries that came out the same week, without a big account behind them, mostly stalled at a few hundred.
He has consistently worked this way: several Indonesian-language AI tool tutorials a week, with 11 of his last 20 videos carrying a Higgsfield promo link. Other libraries on his GitHub, including streamflow (a YouTube livestream management tool, 410 stars), bang-story (an Indonesian-language AI story-to-video app, 195 stars), and bang-film (an AI short-drama skill he released in Oct 31), also come with tutorial videos. The open-source libraries are material and hooks for his videos, and the money comes back from the paid tools those videos promote (inference, based on the promo links in the video descriptions).
Wave two was repost accounts. From Oct 31 to Nov 6, a batch of accounts that specialize in "discovering useful GitHub libraries" posted it in concentrated bursts. Prinon Turio's caption read "Your AI coding agent can make motion, not another boring slideshow... comment MOTION for the repo link," which is basically the README's first sentence reworded. Individual posts didn't get much reach, a few hundred to a thousand views each. I could not find a single post that explains the +103 on Nov 1; my inference is that several posts stacked together. Same-category libraries showed no daily star change during the same period, which suggests this wave was aimed at this project specifically, not a rising category.
It was never featured on Hacker News or Product Hunt. No big X account mentioned it, and there is no dedicated Reddit thread.
Why people wanted it at this moment
Late September to October, "letting AI make videos with code" became one of the hottest topics in developer circles: HeyGen's open-source framework HyperFrames gained thousands of stars every day in October, and Anthropic released Claude Motion, built into Claude, on Oct 8. Tutorials on TikTok for "making motion graphics with Claude" pulled 300k to 630k views each. People who watched them went looking for "skills I can install in my agent," and Bang Motion got picked up by directories and lists like skillleaderboard, claudemap, and awesome-claude-video-skills. People often pair it with HyperFrames. On Oct 14, @ikhlasacodes's Indonesian TikTok on "Claude Code + Bang Motion + HyperFrames" hit 10.7k views.
Who it's competing with
| Project | Launched | Size (11-06) | Notes |
|---|---|---|---|
| Claude Motion(Anthropic) | 10-08 | Built into Claude | Official one-click, hosted inside Claude |
| heygen-com/hyperframes | 2026-03 | 79,455 stars, +4,109 in a week | The underlying framework for "write HTML, render video" |
| remotion-dev/remotion + skills | 2020 / 2026-01 | 71,001 / 6,409 stars | React rendering pipeline and official skill |
| latent-spaces/brag | 2026-06 | 20,413 stars | Turn a finished project into a launch video in one click |
| Vincentwei1021/video-shotcraft and 4 other libraries | 7 ~ October | 17,238 stars (the other three each have 2,000~3,000) | One person put out a whole set, built on Remotion |
| LottieFiles/motion-design-skill | 2026-03 | 2,805 stars | General UI motion principles, doesn't produce finished video |
| Tejashmakwana/motionmaxxing | 10-06 | 1,378 stars | The pitch is nearly identical: turn AI-looking video into cinematic video |
| Barty-Bart/motion-graphics | 9-25 | 1,089 stars | Same time, same scale |
| cinetic, claude-motion, claude-motion-design, motion-graphics-skills | 9-26 ~ 29 | 500 ~ 1,000 stars | A batch that appeared the same week in late September |
Searching for "motion graphic / remotion / hyperframes / explainer" turns up over forty libraries with 700+ stars. Bang Motion ranks mid-pack.
The differentiation is real, but it's not why it took off
Outputting a single HTML file, briefing before coding, and working across four agents are genuinely different from the competition. But every similar project claims "cinematic" and "not a slideshow," and from the outside, nobody can tell whose rules are better just by reading the README. The extra stars it got over peers like Barty-Bart came from the author's channel and good timing. The differentiation mainly made the README look good and the one-liner easy to repeat, which is what the repost accounts liked.
The code hasn't changed since Sep 15. The author's attention has already moved to bang-film. Once this wave of reposts passes, it will most likely settle back to a few dozen stars a day.
If it were you
Verdict: Don't follow the category, but the tactics are worth learning. "Agent motion skills" already has over forty libraries with 700+ stars, led by company projects like HyperFrames and Remotion, with Anthropic itself now in the game. Bang Motion reached a thousand because of a 1.41M-subscriber channel that a follower can't copy.
Good for: Developers who already run a tutorial channel (YouTube, Bilibili, or Douyin) and post on a fixed schedule, or who know a non-English dev community well enough to be the first to cover something there.
Not for: Anyone with no distribution channel who expects stars to come from a good README and clear rules, or anyone who wants to build the 41th generic motion skill.
How to do it: Copy its loop, not its category. Pair every tutorial video with a small open-source library you wrote, put the repo link in the video description, open the README with a pain point (theirs is "AI gives you slideshows"), and use a multi-panel finished-product collage as the cover image. If you really want to build a video skill, pick one narrow niche they haven't covered well: for example, short product-ad videos for 9:16, Chinese subtitles and voiceover sync, or a style-rule layer for one specific rendering framework.
How to get users: Your own videos are the first wave. The second wave comes from repost accounts: the more screenshot-friendly the README and the easier the one-line pitch, the more likely accounts like gittrend.io and GitFinder are to cover it. Their "comment a keyword to get the link" tactic can add a couple hundred stars a day for a few days. Also submit the library to skill directories and awesome lists.
When to move: Within one to two weeks after the next big-company or big-framework launch (the next Claude Motion or a major HyperFrames release). Miss that window and the top spots are taken. The late-September week already showed this.
Signs you've misread it: ① Bang Motion gains another 1,000 stars in a month with no new videos or reposts, which would mean its rules really are being adopted and quality wins; ② a new motion skill with no creator audience launched in November reaches 3,000 stars in a month, which would mean the category still has room; ③ similar libraries mostly stall in November, which would mean the heat has passed and even narrow niches aren't worth it.
Data and sources: GitHub repo page, README (v1.19.0), and daily star counts (10-17 to 11-06); the author's other GitHub repos; YouTube channel Bang Tutorial and the videos "Bikin Motion Graphic Pakai ChatGPT", AutoClaw, Prinon Turio, GitFinder, and Fabiano Carvalho (view count 11-06); Reveneau 2026-09-24 coverage; Instagram @gittrend.io and @synthetix_motion; TikTok @ikhlasacodes and @prinonturio; Claude Motion help doc (10-08); skillleaderboard, claudemap, awesome-claude-video-skills; Hacker News and Product Hunt records; competitor GitHub pages.
Research notes: competitor table, acquisition evidence, full research report
- Competitors
Name Where Launched Scale How it differs from this one Claude Motion(Anthropic) Big-company features 2026-10-08 Built-in /motion in Claude; single related TikTok tutorials at 340k to 600k views Official one-click hosting; bang-motion works across 4 agents, outputs a single HTML file you can take anywhere, and has more detailed style rules heygen-com/hyperframes Open source 2026-03 79,455★, +4,109 this week A low-level HTML-to-video rendering framework; bang-motion is the higher-level design rules and workflow on top, and the two are often used together remotion-dev/remotion + remotion-dev/skills Open source 2020 / 2026-01 71,001★ / 6,409★ A React rendering pipeline; bang-motion doesn't depend on Remotion and plays directly in the browser Vincentwei1021/video-shotcraft and 4 other libraries Open source 2026-07~10 17,238★ (plus 3,140/3,066/2,118★ elsewhere) The same author has a set of libraries about ten times that size, built on Remotion and aimed more at product promo videos latent-spaces/brag Open source 2026-06 20,413★ Turns a finished project into a release video; narrower scope LottieFiles/motion-design-skill Open source 2026-03 2,805★ General UI motion principles, doesn't produce finished video iart-ai/motion-skills Open source 2026-06 1,233★ A collection of 50 skills; bang-motion is a single skill with a brief-first workflow Barty-Bart/motion-graphics Open source 2026-09-25 1,089★ Similar-size Claude Code/Codex motion skills from the same period, with no creator audience behind them Tejashmakwana/motionmaxxing Open source 2026-10-06 1,378★ Nearly the same pitch (turning AI-looking video into cinematic video) - Where users come from
- Verified the author's own YouTube channel (with a Higgsfield affiliate link): Bang Tutorial (1.41M subscribers, Indonesia) 9-19 video "Bikin Motion Graphic Pakai ChatGPT (GPT-6 Astra + Higgsfield MCP)" with 103K views. The description says "my own free open-source skill" and links the repo. 5 days later it has 9-24 with 493★
- Verified Riding platform hype (HyperFrames, Claude Motion): 10-08 Claude Motion launched; single Claude motion-design tutorials on TikTok/YouTube reached hundreds of thousands of views; Indonesian TikTok @ikhlasacodes 10-14 demoed it with Bang Motion + HyperFrames, 10.7k views
- Verified Short-form videos from GitHub repost accounts (TikTok/IG/YT Shorts): gittrend.io 10-31 (TikTok 992 views), @synthetix_motion on IG 11-01, Prinon Turio 11-02 (YT 1,043), GitFinder 11-03, Fabiano Carvalho 11-05, Trending Open Source Projects 11-06; these line up with the jump starting 11-01
- Verified Skill directories and awesome-list SEO: listed on skillleaderboard, agentskillshub, claudemap, claudepluginhub, codexstacks, and zhuyansen/awesome-claude-video-skills
- Inferred Product partnership promotion (AutoClaw by Z.ai): AutoClaw's channel posted 9-16 a video titled "made a 4K promo with the just-open-sourced Bang Motion" that got only 140 views; whether there's a partnership is unverified
bangtutorial/bang-motion: a skill from an Indonesian YouTube tutorial channel with 1.41M subscribers, built to stop AI motion graphics from looking like slideshows
Research date 2026-11-06. Data card: 62 days, 1,094 stars, 140 forks; last 7 days +332 (the prior 7 days +80); single-day peak 11-01 +103.
Bottom line
bang-motion is an agent skill written by Bang Tutorial (an Indonesian YouTube channel with 1.41M subscribers). It first took off thanks to a sponsored Higgsfield video on the author's own channel (9-19.1M views), and rode the late-September-to-October "AI motion graphics" wave (HeyGen's HyperFrames went viral, then Anthropic launched Claude Motion, 10-08). Its real differentiation is writing a motion designer's craft as rules an agent can check itself, plus a workflow that produces a style brief before any code, not rendering tech. But this category is now crowded with dozens of libraries, and it's mid-sized there. The early-November star surge wasn't pushed by the author; a group of "good GitHub repo" repost accounts (gittrend.io, GitFinder, Prinon Turio, Brazil's Fabiano Carvalho, and others) posted its short videos in a cluster between 10-31 and 11-06.
1. What it is
- An Agent Skill (SKILL.md + starter template + scripts) that supports Claude Code (plugin marketplace), Codex, Gemini CLI, and Cursor. The output is a single index.html that plays and loops automatically on double-click.
?debug=1brings up a scrub bar, and you can export to MP4 frame by frame with Puppeteer + ffmpeg, or sync it to a voiceover. - Menu-style design: 10 style directions × 19 title-card structures × 6 explainer styles. The workflow produces a "style brief" first (3 candidate concepts + a recommendation + a color palette with the source of each color), and code is only written after the user agrees. Once it's written, the agent captures keyframes and checks them against the rules itself.
- The README's pitch: "When you ask AI for a video, you usually get slides." It turns "one continuous world, shots that really move, a subject that runs through the whole piece, and text that lives inside the scene" into rules an agent can check in its own code. It also shows building a cartoon explainer directly in After Effects through the Higgsfield MCP.
- Source: github.com/bangtutorial/bang-motion (README, v1.19.0). The public repo has only 3 commits, all in 9-15 (v1.17.0→v1.19.0); earlier history was squashed, and there have been no code pushes since.
2. Who the author is: the key to understanding it
- GitHub account bangtutorial, bio "Content Creator", with only 156 followers on GitHub.
- The account matches the YouTube channel Bang Tutorial (@BangTutorial): based in Indonesia (country=ID), 1.41M subscribers, 589 videos, 189M total views (YouTube Data API, 2026-11-06). The
youtu.betutorial link in the repo description and the video description's "Bang Motion, my own free open-source agent skill" both point to the same person. - His content pattern: several Indonesian-language AI tool tutorials a week, most with Higgsfield referral links (11 of his last 20 videos include a higgsfield.ai/s/…bangtutorial link). The other libraries he's put on GitHub follow the same playbook: streamflow (YouTube live-stream management, 410★), bang-story (desktop app for making Indonesian-language AI story videos, 195★), aether-clash (browser fighting game, 19★, with a tutorial video), and bang-film (10-31, a new AI live-action short-drama skill, 4★).
- In other words: the open-source libraries are the material and hooks for his tutorial videos, and the money comes back through Higgsfield affiliate/sponsorship deals, not from the libraries themselves.
3. Growth timeline (star counts from our github repo_day data, except where noted)
| Date | Event | star |
|---|---|---|
| 09-05 | Repo created | 0 |
| 09-15 | Public v1.17–1.19; AutoClaw (a Z.ai product) channel posted a video "made a 4K promo with the just-open-sourced Bang Motion" on 09-16, with 140 views | — |
| 09-19 | Bang Tutorial posted "Bikin Motion Graphic Pakai ChatGPT (GPT-6 Astra + Higgsfield MCP)", promoting bang-motion in the description and doubling as a Higgsfield promo; 103k views, 1,913 likes | — |
| 09-24 | Reveneau (an AI-generated news site) covered it: "19 days, 493 stars, 72 forks" | 493(Reveneau) |
| 09-25~09-29 | A cluster of similar motion skills appeared: Barty-Bart/motion-graphics, charlie947, Leonxlnx/cinetic, howseen-ai, whaleyxbt/claude-motion, Mort1d, and others (creation dates in the competitor table) | — |
| 10-08 | Anthropic launched Claude Motion (support.claude.com help doc 10-08) | — |
| 10-14 | Indonesian TikTok @ikhlasacodes: "built with Claude Code + the Bang Motion skill + HyperFrames", 10.7k views | — |
| 10-17 → 10-31 | Steady, 7–16 per day | 637 → 778 |
| 10-31 | gittrend.io (a GitHub repo-recommendation account) posted short videos on TikTok/Instagram, 992 TikTok views; Malaysian TikTok @coachkhairul_ck posted "4 skills to install for motion design with Claude Code", 18k views (I didn't see the video content, so whether bang-motion is included is unverified) | 778 |
| 11-01 | Instagram @synthetix_motion posted a "Make explainer videos with Claude Code in 4 steps" tutorial, and step one is installing Bang Motion | +103 → 881 |
| 11-02 | YouTube/TikTok Prinon Turio, "Turn Your AI Coding Agent Into a Motion Designer with Bang Motion" (1,043 YT views, "comment MOTION for the link") | +78 → 959 |
| 11-03 | YouTube Shorts GitFinder (296 views) | +33 |
| 11-05 | Brazilian YouTube Fabiano Carvalho, Portuguese intro (668 views, "comment for the link") | +46 |
| 11-06 | YouTube "Trending Open Source Projects" (same outfit as gittrend.io) Shorts, 857 views | +29 → 1,094 |
Key points:
- The jump in 11-01 is its own, not the whole category's: over the same period, the peer repos (Barty-Bart, charlie947, whaleyxbt, cinetic, howseen-ai) show no change in their daily star gains. Only remotion-dev/skills also had a jump on 11-01 (+166). So it came from specific content that mentioned it. What we can find is the batch of 10-31~11-02 reposted short videos and Instagram tutorials in the table above; each has modest views and none clearly explains the +103 on its own, so we haven't found the source.
- The September launch phase (19 to 493) most likely came mainly from the author's own 100k-view video. Our library 10-17 only started collecting then, and we have just one data point from Reveneau, so we can't verify it day by day.
- Not found in our HN or Product Hunt data; x-radar's X corpus has nothing either (searched bang-motion / bang motion / bangtutorial); no dedicated Reddit threads found.
4. Competitors: the same thing is already crowded
"Have a coding agent make motion graphics/promo videos with code" was one of the most crowded open-source niches from 2026 9 through November. Searching our GitHub library for motion graphic / remotion / hyperframes / explainer and similar keywords turns up over forty repos with ≥700 stars. The main ones:
| Name | Type | Launched | Scale (GitHub repo 11-06) | How it differs from bang-motion |
|---|---|---|---|---|
| Claude Motion(Anthropic) | Big-company features | 2026-10-08 | Claude has built-in /motion; related TikTok tutorials get 340k~600k views each | Official one-click, hosted inside Claude; bang-motion works across 4 agents, outputs a single portable HTML file, and has more detailed style rules |
| heygen-com/hyperframes | Open source (company) | 2026-03 | 79,455★, +4,109 this week | An underlying framework for "writing HTML to render video"; bang-motion is the higher-level design rules and workflow, and people often use the two together |
| remotion-dev/remotion + remotion-dev/skills | Open source (company) | 2020 / 2026-01 | 71,001★ / 6,409★ | React rendering pipeline + official skill; bang-motion doesn't depend on Remotion, uses GSAP/CSS, and plays directly in the browser |
| Vincentwei1021/video-shotcraft (plus mg-styles-15, anything2explainer, video-talkcraft) | Open source | 2026-07~10 | 17,238★ / 3,140★ / 3,066★ / 2,118★ | One author has put out a whole set of repos, roughly a dozen times bang-motion's scale; built on Remotion, leaning toward product promo videos |
| latent-spaces/brag | Open source | 2026-06 | 20,413★ | Turns the project you just finished into a launch video in one click; narrower, more direct scenarios |
| LottieFiles/motion-design-skill | Open source (company) | 2026-03 | 2,805★ | General motion principles (easing, choreography), mainly UI motion, doesn't output finished videos |
| iart-ai/motion-skills | Open source | 2026-06 | 1,233★ | A big collection of 50 skills; bang-motion is a single skill with a strong workflow |
| Barty-Bart/motion-graphics | Open source | 2026-09-25 | 1,089★ | Nearly the same scale and timing, Claude Code/Codex motion skills |
| Tejashmakwana/motionmaxxing | Open source | 2026-10-06 | 1,378★ | Also pitches "turn AI-generated slop video into cinematic film," with nearly identical selling points |
| Leonxlnx/cinetic, charlie947/motion-graphics-skills, whaleyxbt/claude-motion, howseen-ai/claude-motion-design | Open source | 2026-09-26~29 | 1,016 / 576 / 845 / 774★ | All from the same September wave, with "cinematic launch film" and "every frame is code" |
| motion.dev AI Kit | Web/commercial | Not verified | Not verified | Motion (formerly Framer Motion)'s official motion tool for agents, leaning toward UI animation |
Bottom line: it's not short on competitors, and most of them are bigger. It sits mid-pack in its category, with no rendering-tech moat.
5. Why it took off
- The author brought the traffic: an Indonesian AI tutorial channel with 1.41M subscribers; the launch video got 103k views and linked the repo in the description. Peer libraries of the same period without a big-channel backer mostly stalled at a few hundred stars. This is the main reason (evidence: YouTube Data API view counts; 9-24 already had 493★).
- Caught right in the hottest weeks of demand: HyperFrames gained over 1,000 stars in October, 10-08 Claude Motion launched, and "make motion graphics with Claude" tutorials on TikTok/YouTube are racking up hundreds of thousands of views each (kaye.creatives 630k, Mr omi 630k, @motionlovee 610k, @nocode.joshua 340k). Viewers who finish watching go looking for a "skill you can install," and bang-motion is the one that shows up in search results and awesome lists (skillleaderboard, agentskillshub, claudemap and zhuyansen/awesome-claude-video-skills all list it).
- The README is itself an ad reel: the header image is a collage of 12 styles, each with a finished frame, and the line "AI gives you slides, not video" names the pain point directly. The repost accounts' copy is basically that sentence word for word ("real motion graphics, not slide decks"). That quotable hook is why the November repost accounts were willing to push it.
- The November wave came from second-hand spread by GitHub repost accounts: gittrend.io (posted to TikTok/IG/YT at once), GitFinder, Prinon Turio, Fabiano Carvalho, all using the "comment a keyword to get the link" DM-funnel tactic. The author didn't arrange any of it; these accounts saw stars climbing and content that was easy to film, so they reposted it.
6. What's the differentiation, and is it the reason it took off?
- Compared with peers, it differs in three ways: (a) it runs only in the browser and outputs a single HTML file, so you can view it without installing Remotion or Node; (b) "design rules + a brief before code", which emphasizes avoiding cookie-cutter output (style derived from the brand, structure chosen from a menu, no copying templates); (c) works across 4 agents, and connects to Higgsfield MCP for generated assets and voiceover.
- These differences are real, but they are not the main reason it took off: many peers in the same period also pitched "not like slides" and "cinematic," and outsiders can't tell whose rules are better. It gained more stars than its peers because of the author's channel traffic and timing with the platform hype; the differentiation's role was to make the README and demo images attractive and quotable, which made repost accounts willing to spread it.
7. How many people have seen it
- Word of mouth in small circles: no coverage on HN, Product Hunt, major X accounts, or mainstream media. What exists is the author's own Indonesian-language videos, Reveneau's AI-generated script, a batch of repost shorts with a few hundred to a thousand views each, and skill directories that auto-indexed it.
- No revenue involved (open source, MIT); the author's income comes from Higgsfield promotion, via a public affiliate link.
- Copycats: none specifically found copying it, but the "agent motion skill" niche itself is saturated (9-25~9-29 saw more than 6 similar libraries in a single week; Vincentwei1021 alone put out 4, the largest at 17k★).
8. What's worth borrowing
- The loop of "tutorial channel + your own open-source tool + affiliate link": the open-source repo supplies footage and hooks for videos, and money comes back from a paid tool promoted in those videos (here, Higgsfield). Anyone with a channel can attach a self-made open-source repo to each video, and the repo in turn drives subscribers back to the channel.
- A one-line pain point + a multi-style collage in the README spreads well; repost accounts will copy your copy word for word.
- Ride the big-company launches: once Claude Motion and HyperFrames created the demand, "a skill that installs into my agent" became a niche worth claiming, provided someone gets it in front of viewers.
9. Still unclear
- We could not find which content drove the 11-01 +103 gain (candidates: gittrend.io 10-31, @synthetix_motion 11-01, Malaysian @coachkhairul_ck 10-31; whether the last one mentioned it is unverified). Stargazer details can't be pulled (API rate limits + 403), so we couldn't check the timing or account distribution of stars.
- No day-by-day curve for the September launch phase (our library 10-17 only started collecting then).
- No code updates since 9-15; the author's attention has shifted to newer repos like bang-film. Once the repost accounts' buzz fades, growth will likely return to a dozen or so stars a day.
- Whether the AutoClaw (Z.ai) video has any partnership with the author is unverified.
Sources
- Repo and README: github.com/bangtutorial/bang-motion; the author's other repos: GitHub API users/bangtutorial/repos (2026-11-06)
- Our GitHub library: repo_profile, repo_day (2026-10-17~11-06 daily star counts; peer comparison)
- YouTube Data API (2026-11-06): channel UC7iH5kaCViSPybhwCIfT1Jg; videos 0ANV5dfL_yQ, qL1ZJY69kc4, swSlZpeSk0w, y4pqyhEWqBQ, QC-wjLJs7oA, LP9R2gavh9c, AsvN_feasyY, OtbLL6OBNE4
- Reveneau 2026-09-24:reveneau.com/ainews/bang-motion-agent-skill-html-motion-grap…
- Instagram: www.instagram.com/reel/DfKJNWvjPgg/ (gittrend.io), www.instagram.com/reel/DfOowvRR45y/ (synthetix_motion, 11-01)
- TikTok (TikHub search): @gittrend.io 10-31, @prinonturio 11-02, @ikhlasacodes 10-14, @coachkhairul_ck 10-31, @nocode.joshua 10-09
- Claude Motion: support.claude.com/en/articles/17454997-get-started-with-cla… (10-08); coverage by howtogeek
- Directory sites: skillleaderboard.com, agentskillshub.top, claudemap.org, github.com/zhuyansen/awesome-claude-video-skills
- Our hackernews and producthunt data: no relevant records; x-radar: no relevant records