# The 740-star open-source project that never had a single day of breakout: Adu's AI video-editing Skill, and the "let AI make video" space where 400+ libraries are competing · GitHub open source

> Source: https://stringers.io/en/github/gh_adunext__adu-motion-video.html (Stringers / 什么值得做, generated 2026-11-08)
> Stringers answers one question every day: what small product can one or two people build in a few months and actually make money from. Numbers come from 15 databases we collect daily, each with source and date.
> For agents: these verdicts and the underlying databases (verified revenue, App Store / Google Play charts, ad libraries, GitHub, Product Hunt…) are queryable live over MCP — tools today / hot / products / directions / direction / data. Setup: https://github.com/komako-workshop/stringers (the user logs in at stringers.io → "配置 agent" to get a personal URL). Site index: https://stringers.io/llms.txt

> Riding the "AI makes video with code" wave after Opus 5.5: Chinese-language, free, and every style can be seen as a finished video up front. The stars came from two X Chinese repost accounts reposting the same text, Xiaohongshu roundups, and directory sites sending a little traffic.

[GitHub open-source project](https://stringers.io/en/github/index.md) · written 11-6

adunext/adu-motion-video · adunext

# The 740-star open-source project that never had a single day of breakout: Adu's AI video-editing Skill, and the "let AI make video" space where 400+ libraries are competing

**What it is** A pack of explainer and talking-head video style templates for Codex / Claude Code. It turns your voiceover, script, and footage into an MP4 with animated captions, plus an editable project.

**Why it took off** It rode the "AI makes video with code" wave after Opus 5.5. It's in Chinese, free, and every style can be watched as a finished video right away. The stars came in gradually from two Chinese X accounts reposting the same post, a Xiaohongshu roundup, and directory sites.

**C** **How you can play this** Don't build the 428th talking-head template library. Study its funnel (open source to build trust → README links to their own platform → then to an app), or build vertical style packs for niches nobody has covered well, like no-presenter videos.

C grade: within two months the category spawned 400+ similar projects, the most direct rival came earlier and has 3× its stars, and this one has leveled off at 740 and is cooling. The playbook is worth learning, but copying it won't get you many stars. Circulating in small circles

Data (updated daily) · Repo created 2026-09-28 · 761 stars · 7 days +106

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

A personal account created 2013 years ago, with only 70 followers, turned a set of talking-head motion-graphics templates into a Skill for Codex / Claude Code and reached about 740 stars in 39 days. It has never had a breakout day, at most thirty to forty stars in a day, and the stars came in gradually from Chinese AI-community reposters, roundup posts, and directory sites. The category already has 400+ similar projects, and this one sits mid-pack. Below: how it gathered stars, what the author actually wants, and how much room is left if you get in now.

### a 2-min 49-sec Bilibili video

On 2026 Sep 29, around noon, the Bilibili account 阿杜 posted a 2 min 49 sec video titled "I 'distilled' Claude's editing ability, the Skill is at the end of the video." As of 11-06, the video has about 2,081 views and only 1 danmaku. The day before, on Sep 28, the repo adunext/adu-motion-video was created. On Oct 1, the author posted on Xiaohongshu under 阿杜 Next a post titled "The first open-source person to publicly replicate Claude's editing ability 阿杜…", with 34 likes and 51 saves. The same-titled video on YouTube has 54 views, and the channel has 26 subscribers.

The numbers are small, but the timing lines up: the repo gained 0 stars a day in its first three days, and growth picked up again in Oct 2.

### What it is

A pack of "style templates." You install Codex or Claude Code, hand it the template, then give it your voiceover recording, script, and footage. It arranges shots, adds animation, captions, and sound effects according to the template, and outputs an MP4 and a project you can keep editing. The README lists 6 styles: modern graphic talking-head, paper-style perspective explainer, dark 3D, dark tech console, bright character narrative, and more. Each comes with a GIF and an online playback page, so you can click in and see what the finished video looks like. It's in Chinese, free, and MIT-licensed.

How it differs from HeyGen's hyperframes (79k stars) and Remotion's official skills (6.4k stars): those are frameworks for "writing video in code," and what the output looks like is up to you to design. Adu gives you a set of ready-tuned looks; the README says it is "not a prompt, not a reference video." Animation, layout, timeline, transitions, and sound effects are all included.

### Where the stars came from: two identical recommendations

Daily GitHub stars (trendshift and GitHub, as of 11-06):

| Timeline | What happened | star |
|---|---|---|
| 9-28 to 10-01 | Repo created; the author posts on Bilibili and Xiaohongshu | 0 per day |
| 10-03 to 10-05 | First small peak | +30+ per day |
| 10-17 |  | 356 |
| 10-26 to 10-28 | Second small peak | +28 to +47 per day |
| 11-06 |  | 731 (GitHub page shows 740), 81 forks |

Over the last 7 days it gained 109 stars; the 7 days before that it gained 163. It's cooling.

Two spikes line up with two X posts. On Oct 5, the Chinese GitHub recommender Geek Lite (@QingQ77) posted: "Adding motion graphics to talking-head videos means writing templates from scratch; this Skill hands ready-made animation scenes to AI, which cuts them into an MP4 plus an editable project to your script and footage." A reply under it said: "Keeping the editable project is what matters, so small text tweaks don't mean redoing everything." On Oct 25, another account, DKing (@dking_ai), posted **the exact same text word for word** , and the second spike follows right after. Identical copy suggests a repost account pushing "recommended repos" in bulk, not someone who genuinely vouches for it.

The remaining ten or so stars a day come from the long tail:

- The Xiaohongshu compilation post "40 open-source projects that let AI edit videos for you" by "AI进化论.斯先生" 10-28 tested it by name, with 23 likes. Bigger rankings, like AI萧船长's "Codex video-making, Skill leaderboard" with 709 likes, did not include it.

- A Traditional-Chinese account on Threads relayed it, opening with "What outsourced talking-head editors most don't want you to know…", with 13 likes.

- Auto-generated write-ups on CSDN and GitCode, several GitHub directory sites, and an awesome-claude-video-skills list that includes 427 libraries in the same category.

No media coverage, no big-account reposts. It's "being passed around in a small circle."

One number shows these stars aren't all casual clicks: the three releases add up to about 3,500 downloads (GitHub, 11-06), four to five times the star count. But the issues contain only 2 real questions, such as "Is there a template that doesn't need a real person on camera?" and "Why is there a duration limit?" There is one contributor, the author, with 53 commits, and the commits carry Co-Authored-By Claude trailers.

### The author is actually paving the way for something else

Open the repo now and the first screen of the README isn't usage instructions, it's a bold line: "Download the latest templates at MotionGit →". From mid-October onward, all templates were converted to MotionGit's `.motion` format, and downloads go through motiongit.com/@adunext.

Looking further out:

- A GitHub account named motiontemplates, registered 10-12, holds 53 repos. Each one is a description page for a `.motion` template, all pointing back to MotionGit. This is planting entry points for GitHub search.

- adudir.com is the desktop and mobile app of "阿杜导演" (ADu Director). The MotionGit editor page reads: "ADu Director brings .motion templates into a professional native editing workspace."

- The repo shipped 6 releases in early October, and after 10-19 there were no more commits. The author's energy moved to the platform.

I couldn't find company registration records showing whether MotionGit belongs to the author, but the copy is strongly similar (inferred). So the repo's role is clear: **use open-source stars to earn trust, then route people to your own template platform and app** . Stars are the entry point, not the goal.

### Who it's competing with

This track heated up only after Claude Opus 5.5 launched. "Let coding agents make videos with code" became a topic in Chinese AI circles in the autumn of 2026. awesome-claude-video-skills was created in Sep 27 and collected 427 libraries in two months, even adding a dedicated section "Made with Claude Opus 5.5".

The closest peers to adu (GitHub, 11-06):

| Project | Launched | Scale | How adu differs |
|---|---|---|---|
| Vincentwei1021/video-talkcraft | 2026-08-22 | 2.1k star | The most direct rival: it also turns a talking-head script plus voiceover into a motion video, launched a month earlier, with 3× adu's stars |
| nateherkai/herk-edit-kit | 2026-04 | 2.6k star | Picks 406 motion cards to match the script, same idea, leaning toward English vertical video |
| Vincentwei1021/anything2explainer | 2026-09 | 3.1k star | Gives it a topic and produces an explainer video directly, with no real person on camera, which answers the question raised in adu's issues |
| Gitruck/cli | 2026-06 | 962 star | Chinese, "raw talking-head footage → an editable finished project", same selling point |
| motion-video-kit, claude-motion-design, bang-motion, motion-skills | 2026-06 to 09 | 0.8k–1.4k stars each | Code-level motion skills of similar scale |
| heygen-com/hyperframes | 2026-03 | 79k stars, +4,109 per week | A big-company general framework with no style |
| Jianying (CapCut CN), Duojia, Xiaohongshu OpenStoryline | Available / in closed beta | Massive user base | Usable with a few taps in the app, no need to run Claude Code |

Lining up roughly 40 top-starred libraries in the track, adu ranks around #35. It isn't the first and holds no exclusive position: the idea of "style templates + agent selection" was already in video-talkcraft and herk-edit-kit, both earlier. What it stands on is two things: Chinese language, and taste.

### If it were you

**Verdict** : C. There's something to learn, but copying it isn't worth much. In two months the track produced 400+ libraries, the closest rival came earlier and is three times bigger, and adu itself has only 740 stars and is cooling off. Building another "talking-head motion template Skill" will most likely be the 428th. What's worth learning is its playbook, not its subject.

**Who it suits** : people who already make motion graphics and have a taste of their own (designers, editors, creators of explainer videos) with a "look" others want; or people who already have a site or portfolio and want an open-source project to send people there.

**Who it doesn't suit** : people who don't know motion design and plan to have AI generate a few templates and list them (exactly what those 400 libraries are doing, and nobody bookmarks them); people who want to become famous from stars alone (the top of this niche is big companies and a few early entrants, and the mid-tier ceiling is a thousand or two stars).

**How to do it** , ranked by how worthwhile:

1. **Copy only its funnel** : open-source something whose finished output you can see directly, build reputation with stars, and put a README whose first screen sends people to your own site or profile; open a separate account, one template per repo, to cover GitHub search. This works in other niches too.

2. **Fill a gap none of them has filled** : educational videos with no real person on camera (people are asking in the issues), or a style pack for one vertical (courses, product sales, finance explainers). Go deep on one, don't cover everything.

3. **Build in another language** : if you can build style packs for talking-head creators in Japanese, Korean or other smaller languages, where English and Chinese are already crowded, that space may still be open (inferred; I haven't checked each one).

**How to get users** : follow adu's path, which costs the least: post your own 2–3 minute demo video, "I turned Claude's editing ability into a Skill"; attach a GIF and an online playback page to each style so people see the finished output at a glance; and proactively send the repo to Chinese GitHub recommenders on X and the authors of Xiaohongshu "N open-source projects" compilations. Each recommendation brings roughly 100 stars (estimated from adu's two spikes). To get more, you'd need to get into the top compilations, and adu stalled at 700 because it never made it in.

**When to move in** : judge this wave of "AI makes videos with code" by the weekly star gains of hyperframes and video-shotcraft. If you want to ride it, ship before year end; once apps like Jianying or Duojia make the same capability a button, talking-head creators outside the developer circle won't need a Skill (inferred).

**What would show this is wrong** : ① awesome-claude-video-skills' listing count keeps rising fast, and new libraries in the talking-head category pass 3,000 stars within two months, which would mean the ceiling here is higher than estimated and it's worth pursuing harder; ② if a direct rival like video-talkcraft drops to single-digit weekly star gains after December, the wave is over, and switching languages or verticals won't earn much either; ③ if Jianying or Duojia officially launches "automatic motion graphics from talking-head footage," this track closes to people outside the developer circle.

### Data and sources

GitHub: the adunext/adu-motion-video repo page, README, release download counts, issues and commit history (11-06); daily stars from trendshift.io/repositories/275777 and GitHub (10-17 to 11-06); star counts for peer repos (11-06). X: @QingQ77 2026-10-05 post and replies, @dking_ai 2026-10-25 post. Bilibili BV16aaH6kEKC (2026-09-29), YouTube @adunext. Xiaohongshu: author "阿杜 Next" 10-01 notes, "AI进化论.斯先生" 10-28 compilation, AI萧船长 leaderboard. Threads @effusioncreatives. awesome-claude-video-skills list. motiongit.com, adudir.com, adunext.com, GitHub user motiontemplates. Where the text says "inferred" or "estimated", it's an inference or an estimate.

[GitHub repo ↗](https://github.com/adunext/adu-motion-video)

**Research notes: competitor table, acquisition evidence, full research report**

Competitors

| Name | Where | Launched | Scale | How it differs from this one |
|---|---|---|---|---|
| heygen-com/hyperframes | Open source | 2026-03 | 79k stars, +4109 this week | A general HTML-to-video rendering framework with no ready-made styles; adu ships ready-made talking-head style templates |
| browser-use/video-use | Open source | 2026-04 | 30.6k stars | Rough cutting of talking-head footage and transcript-based editing, without motion packaging |
| Vincentwei1021/video-shotcraft | Open source | 2026-07 | 17.2k stars | Uses Remotion for cinematic product promo videos; adu focuses on talking heads and explainers |
| Vincentwei1021/video-talkcraft | Open source | 2026-08-22 | 2.1k stars | The most direct rival: it also turns a talking-head script plus voiceover into a motion video, launched a month earlier, with 3× adu's stars; adu adds a narrative of "open-format templates plus a platform" |
| nateherkai/herk-edit-kit | Open source | 2026-04 | 2.6k stars | Picks 406 motion cards to match the script, leaning toward English vertical short video |
| remotion-dev/skills | Open source | 2026-01 | 6.4k stars | Remotion's official skill; you bring your own design |
| Gitruck/cli | Open source | 2026-06 | 962 stars | Chinese "raw talking-head footage → an editable finished project", more about editing workflow than style templates |
| Jianying / Duojia / Xiaohongshu OpenStoryline | Big-company features | Available / in closed beta | Massive user base | Works inside the app, no coding agent or environment setup needed; adu is for people who use Claude Code/Codex |

Where users come from

- **Confirmed** **The author posts videos themselves (Bilibili / YouTube / Xiaohongshu)** : Bilibili BV16aaH6kEKC (9-29, roughly 2081 plays); Xiaohongshu "阿杜 Next" 10-01 post (34 likes, 51 saves); matches when stars began rising on 10-02

- **Confirmed** **Chinese GitHub recommenders on X** : @QingQ77 posted on 10-05 and @dking_ai posted the identical recommendation on 10-25, matching the star spikes at 10-03~05 and 10-26~28 respectively

- **Confirmed** **Xiaohongshu compilation post** : "40 open-source projects that let AI edit videos for you" by "AI进化论.斯先生" (10-28,23 likes) tested it by name; the top-tier leaderboard (AI萧船长, 709 likes) did not include it

- **Confirmed** **SEO / directory sites / awesome lists** : auto-generated articles on CSDN/GitCode, HuanHQ review, uithub, gitrank, skillsllm, nicheloom, and inclusion in awesome-claude-video-skills

- **Confirmed** **Threads repost** : @effusioncreatives posted, 13 likes

- Inferred **WeChat Official Account / Douyin** : found a WeChat article link that wouldn't open; nothing found on Douyin

## Why adunext/adu-motion-video (adu's AI editing Skill) gained traction

> Channel: GitHub · research day 2026-11-06 · author Michael Du (GitHub account registered 2013 years ago, 70 followers, personal account)

### One line

A motion-template pack for "talking-head + explainer" videos, for use with Codex / Claude Code and similar tools: the assistant reads the templates and arranges your script, copy and footage into a video with animation, subtitles and sound effects, delivering an MP4 and an editable project. It gained about 740 stars over 39 days since it was created, but **it never had a single breakout day** . The stars accumulated through secondary spread in Chinese AI tool circles: Chinese GitHub recommenders on X, Xiaohongshu compilation posts, auto-generated CSDN/GitCode articles, and awesome lists. It rides the wave of "Claude Opus 5.5 can make videos with code", and the track already has 400+ libraries. For the author, the repo is really **a traffic funnel into their own MotionGit template platform and the 阿杜导演 desktop app** .

### 1. Data

- Stars: 2026-11-06 is 731 (our database) / 740 (GitHub page), forks 81–82, fork-to-star ratio about 11%, roughly the median for new repos of similar size, with no sign of star farming.

- Trendshift (trendshift.io/repositories/275777): tracked from 9-28, 9-29 to 10-01 at 0 per day; **stars started climbing from 10-02** , with 10-03 +35, 10-05 +34, 10-12 +30, 10-15 +34, 10-26 +34, and 10-27/28 each +28, then it settled to 9–22 per day. The peak day was only 30 stars (our snapshots show 10-28 as +47, different counting). Last 7 days: +109; the 7 days before that: +163. It's cooling down.

- Release downloads (GitHub API): v3.14.0-rc.1 has 1995, v3.14.1-rc.1 has 1354, v3.3.0 has 143, about 3500 total. **Downloads run several times the star count** , so people are actually installing it, not just starring it.

- Issues: only 2 real user questions ("is there a template without a live presenter", "why is there a duration limit"), so community engagement is shallow. Only one contributor, the author, with 53 commits, and commit messages carry Co-Authored-By Claude.

- Trending: the repo has been up for under 60 days, and by pattern it wouldn't make trending. Trendshift has never had it in its top 25.

### 2. Competitors (the niche is already crowded)

awesome-claude-video-skills (1652 stars, created 2026-09-27) lists **427** libraries for "coding agents make video", and adu-motion-video is among them (the 730-star row). Main rivals below; star counts come from our GitHub database, 2026-11-06:

| Product | Type | Launched | Scale | How it differs from adu |
|---|---|---|---|---|
| heygen-com/hyperframes | Open source (big vendor) | 2026-03 | 79k stars, +4109 this week | A general "write HTML, render video" framework with no ready-made style; adu sells ready-made style templates |
| browser-use/video-use | Open source | 2026-04 | 30.6k stars | Does editing (transcribing talking-head footage, rough cuts), not motion packaging |
| Vincentwei1021/video-shotcraft | Open source | 2026-07 | 17.2k stars, +2584 this week | Uses Remotion for cinematic product promo videos, with 124 shot cards; adu focuses on talking-head and explainer content |
| remotion-dev/skills | Open source (official) | 2026-01 | 6.4k stars | Official Remotion skill; output looks close to default, so you bring your own design |
| Vincentwei1021/video-talkcraft | Open source | 2026-08-22 | 2.1k stars | **The most direct rival** : same idea of turning a script + voiceover into a motion talking-head video, shipped a month earlier, with 3 times adu's stars |
| nateherkai/herk-edit-kit | Open source (YouTuber Nate Herk) | 2026-04 | 2.6k stars | Vertical short-video pack: 406 motion cards chosen to match the script. Same idea: template library + agent picks |
| Vincentwei1021/anything2explainer | Open source | 2026-09 | 3.1k stars | Turns a topic into an explainer video directly, no live presenter needed |
| Agentchengfeng/chengfeng-videocut-skills | open source (Chinese) | 2026-01 | 3.2k stars | Chinese editing agent, more about cutting than motion |
| echris6/motion-video-kit, howseen-ai/claude-motion-design, bangtutorial/bang-motion, iart-ai/motion-skills | Open source | 2026-06~09 | each 0.8k–1.4k stars | All code-level motion skills, same scale as adu |
| Gitruck/cli | open source (Chinese) | 2026-06 | 962 stars | "Talking-head raw footage → re-editable cut project", same selling point |
| CapCut / Dujian (Baidu) / Xiaohongshu OpenStoryline | Big-company features | Available / in closed beta | Massive user base | Works with a few clicks in an app, no environment setup; adu requires knowing Codex/Claude Code |

Bottom line: the niche isn't short on similar projects. adu isn't first, and its star count sits mid-pack (roughly ranked 35 of the 40 repos in the table above). It has no exclusive position.

### 3. Where users came from (growth tactics)

1. **The author's own videos (the initial spark)** : Bilibili "I 'distilled' Claude's editing skill, the Skill is at the end credits" (BV16aaH6kEKC, 2026-09-29, about 2081 views); YouTube @adunext video with the same title (54 views, channel has 26 subscribers); Xiaohongshu "阿杜 Next" post "the first person to openly replicate Claude's editing ability..." (2026-10-01,34 likes, 51 saves). The timing lines up with stars rising from 10-02, but the volume is small; this was only the spark.

2. **Chinese GitHub recommendation accounts on X reposting (both peaks line up)** : Geek Lite @QingQ77 posted on 2026-10-05, matching the first peak from 10-03~10-05; DKing @dking_ai posted **the exact same copy word for word** , matching the second peak from 10-26~10-28 (2026-10-25). Identical text across both suggests batch "repo recommendation" reposting accounts, not genuine endorsements by creators.

3. **Threads repost** : @effusioncreatives (traditional Chinese) posted "The thing talking-head editing outsourcers least want you to know...", 13 likes.

4. **Xiaohongshu roundup post** : "AI进化论.斯先生"'s "40 open-source projects that let AI edit videos for you" (2026-10-28,23 likes) names and tests adu-motion-video. Bigger roundups didn't include it, such as "AI萧船长"'s "Codex makes video, Skill ranking" (709 likes) and "拨云开源研究所"'s list. **It didn't get top-tier traffic on Xiaohongshu** .

5. **SEO and content farms** : auto-generated CSDN/GitCode articles ("adu-motion-video lib.js API full reference", "What is the .motion open format"), a HuanHQ blog review (2026-10-25), and directory sites like uithub, gitrank, nicheloom, skillsllm, plus the awesome-claude-video-skills list. This is a steady long-tail source that matches the flat ten-to-twenty stars a day in the second half.

6. **Author's network and platform funneling** : the author turned the repo into a MotionGit landing page, with "Go to MotionGit to download" on the first screen of the README. The GitHub user motiontemplates (2026-10-12 followers) owns 53 repos, each a .motion template description page linking to motiongit.com/@adunext and similar, forming the platform's GitHub SEO network. adunext.com and adudir.com (ADu Director desktop and mobile apps), and the MotionGit Editor page, say "ADu Director brings .motion templates into a professional native editing workspace", with asset CDN on openaicut.com. I **found no business registration or copyright info** confirming whether MotionGit and the author are the same team, but the copy is clearly from the same source.

### 4. Why it got to 700 stars

- **Timing** : after Claude Opus 5.5 launched, "AI makes video with code" became a hot topic in Chinese AI circles in autumn 2026 (hyperframes +4k stars/week, video-shotcraft +2.6k stars/week). Any repo carrying a "Claude/Codex editing Skill" label picked up some traffic, and the awesome list even has a "Made with Claude Opus 5.5" section.

- **Targets talking-head creators** : the pitch is concrete, "want motion graphics for talking-head video without building templates from scratch", with 6 style GIF previews and an online playback page (adunext.github.io gallery). You can see what the finished product looks like at a glance, and the barrier is lower than frameworks like hyperframes.

- **Chinese, free, MIT** : domestic talking-head and knowledge creators are the main audience. The copy and onboarding are all Chinese, and it includes compatibility notes for Doubao, DeepSeek, and WorkBuddy.

- **Reposted in bulk by batch accounts and SEO sites** : no big accounts; it relied on two X recommendation accounts reposting it, plus long-tail directory sites, adding a dozen-odd stars a day.

- **Ongoing iteration by the author** : released 6 versions in early October, then shifted to MotionGit around 10-18. No new commits since 10-19; growth is coasting on existing exposure.

### 5. Differentiation

- It claims its edge is that it **packages a "style" as a complete project template** : animation source, timeline, transitions, sound design, and replacement notes all live in one `.motion` file. Any assistant can reproduce the same style from it, and the output is an editable project you can build the next episode on. Most rivals give you a framework (hyperframes, Remotion) or an editing workflow (video-use).

- But the edge is **not exclusive** : video-talkcraft (a month earlier, with 3 times the stars), herk-edit-kit (406 motion cards), and Gitruck/cli all use the "template/card + agent picks" approach. adu's distinctiveness lies in its aesthetics (6 styles, refined by the author and Opus 5.5) and the "open format" story.

- **Did the differentiation drive the growth?** Only partly. Star count was mainly driven by timing plus Chinese-language reposting. The ".motion open format" was added in mid-October, when growth was already underway, so it reads more like the author's business positioning for his own platform than a reason for stars.

### 6. Visibility and copycats

- Visibility: **circulating in small circles** . No media coverage; on X it's just two accounts reposting the same text, not in the big Xiaohongshu roundups, and the author's platforms total a few thousand views. Revenue isn't public. It's a free MIT project; monetization runs through MotionGit and the ADu Director app.

- Copycats: no obvious ones yet. Forks show only a few renamed ones (e.g., Jackywxsz/Jacky-motion-video, 1 stars). The niche itself is very homogeneous, but those repos appeared at the same time; they aren't copying it.

### 7. What others can learn

- The funnel "open-source Skill → own template platform → desktop/mobile app": GitHub stars build trust and SEO, the README's first screen drives traffic to the author's site, and a GitHub account (motiontemplates, 53 repos) builds a long-tail SEO template library.

- Preview first: each style comes with a GIF and an online playback page, so you see what the finished product looks like instantly. For video skills, this is the conversion key.

- Actively works with Chinese GitHub recommendation accounts and Xiaohongshu roundup authors, getting roughly 100 stars per wave at low cost.

### 8. Still unclear

- 10-02 The exact source of the first wave of stars is unclear: possibly Bilibili or Xiaohongshu videos, or a WeChat Official Account post I couldn't access (I found one mp.weixin link, but it wouldn't open).

- Reach and engagement numbers for the two X repost accounts (not in the x-radar local corpus).

- Whether MotionGit and openaicut.com are run by the author or his team, and whether the "Coming soon" industry templates on MotionGit will be paid.

- Douyin spread wasn't checked (the TikHub Douyin search endpoint didn't work).

### Sources

- [github.com/adunext/adu-motion-video](https://github.com/adunext/adu-motion-video) (README, releases, issues, commits via the GitHub API)

- [trendshift.io/repositories/275777](https://trendshift.io/repositories/275777) (daily star growth)

- [x.com/QingQ77/status/2107017445886304484](https://x.com/QingQ77/status/2107017445886304484) (2026-10-05), [x.com/dking_ai/status/2114285503134523511](https://x.com/dking_ai/status/2114285503134523511) (2026-10-25)

- [www.bilibili.com/video/BV16aaH6kEKC/](https://www.bilibili.com/video/BV16aaH6kEKC/), [www.youtube.com/@adunext/videos](https://www.youtube.com/@adunext/videos)

- Xiaohongshu notes 6abe7596000000001b02c4cb (author), 6ae1e586000000002b03f642 (40 open-source projects), 6ae4a3ec00000000260245ae (AI萧船长 ranking, does not include adu) (TikHub)

- [www.threads.com/@effusioncreatives/post/DeDx6Kak0TR](https://www.threads.com/@effusioncreatives/post/DeDx6Kak0TR), [www.huanhq.com/blog/adu-motion-video](https://www.huanhq.com/blog/adu-motion-video), CSDN gitblog articles

- [github.com/zhuyansen/awesome-claude-video-skills](https://github.com/zhuyansen/awesome-claude-video-skills)

- [motiongit.com/@adunext](https://motiongit.com/@adunext), [motiongit.com/](https://motiongit.com/), [www.adudir.com/](https://www.adudir.com/), [www.adunext.com/](https://www.adunext.com/), GitHub user motiontemplates

- Competitor star counts: agent-im server github repo_profile, 2026-11-06

This article was written by AI from public data and the data we collect daily. Numbers are sourced, and inferences are labeled.