# How Yingce climbed to 1,675 stars in a crowded AI short-drama open-source category, ranking in the teens · GitHub open-source project

> Source: https://stringers.io/en/github/gh_ddcat-ai__open-ai-canvas.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

> It caught the 2026 AI short-drama wave, shipping updates almost daily, and collected stars piecemeal through Chinese X repost accounts and roundup articles. The MIT license plus a ready-made paid backend made people willing to fork it and API vendors willing to sponsor it.

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

ddcat-ai/open-ai-canvas · ddcat-ai

# How Yingce climbed to 1,675 stars in a crowded AI short-drama category, ranking in the teens

**What it is** An open-source AI short-drama workbench: drop in a novel and it automatically generates scripts, characters, storyboards and finished videos, with a built-in operations backend for credits, payments and model pricing.

**Why it took off** It rode the 2026 AI short-drama boom, shipping an update almost every day, and picked up stars from scattered Chinese-language X repost accounts and roundup posts; the MIT license plus a ready-made paid backend made people willing to fork it, and API vendors willing to sponsor it.

**C** **How to play it:** Stop building yet another short-drama workbench. Take its playbook (open-source core, a back office that can charge right away, sponsor slots) and apply it to generation areas that aren't crowded yet (e-commerce video, audiobooks, comics). Give it a name that's easy to search, and launch through Chinese repost accounts first.

Grade C: this category already has two heavyweights with over ten thousand stars and a dozen-plus projects with several thousand stars. Yingce updates this often and still ranks only in the teens, so copying it won't earn stars. What's worth learning is its monetization structure, not the project itself. Being passed around in small circles

Data (updated daily) · Repo created 2026-07-22 · 1,710 stars · 7 days +126

GitHub · stars

Data 2026-10-09 → 2026-11-08 (31 days, 24 observations). Points before 10-28 come from deeply / archive backfill, not every day has one

Yingce (ddcat-ai/open-ai-canvas) is an open-source AI short-drama workbench: drop in a novel and get storyboards and finished video. The repo is 107 days old with 1,675 stars, ranking in the teens among its peers, which doesn't count as breaking out. The more telling numbers are 545 forks and the 10 sponsors listed in the README, most of them selling model API access. The way it's being used is that people take it and run paid sites with it.

### A launch post with 251 views and a repost with 12K views

On Jul 2026 22, the day the repo was created, X user @liangwenhao3 posted with nothing but two hashtags and the repo description: "#AI #InfiniteCanvas An open-source infinite-canvas workbench for AI film and video creation," which got 251 views. Over the next few days, Chinese tech repost accounts like Geek Lite (7-30) and @vintcessun (8-01) passed it along. By early September it had gathered roughly 800 stars (estimated by working back from the Trendshift number after 9-07).

Three months later, in Oct 29, the account "知识分享官" posted a four-screenshot introduction: "A practical open-source AI film workbench: Yingce, built for people making AI short dramas and videos… bringing the whole creative pipeline into one tool." It drew 12K views and 113 likes. The next day the repo gained +44 stars, and the day after +77, its biggest day on record. The timing matches, but whether this post caused it can't be confirmed (inference).

That's its growth: it never made GitHub trending, no major account wrote about it, no HN, just Chinese X repost accounts, personal blogs, and CSDN and Sohu roundup articles, adding up bit by bit.

### What it is

The README's first paragraph is clear: upload a novel, the pipeline adapts it, splits it into episodes, builds characters, produces storyboards, then generates video segment by segment and stitches the final cut; from segment 2 onward it automatically uses the previous segment's last frame as the first frame so shots connect. There's also an infinite canvas where images, video, text, audio and 3D previsualization are all nodes, and an Agent on the canvas can act directly; finally there's a timeline editor that can export CapCut drafts. The next version is in internal testing as a "World Engine," where you place people and camera positions in a 3D scene and say in one sentence for a character to "run from point B to point A."

Technically it's a React + Bun frontend with a Go backend, SQLite locally, PostgreSQL + Redis for deployment, and a one-click Docker install. The demo site yingce.tv publishes test credentials (test / test123456), so you can try it without installing.

It updates very fast: the CHANGELOG records 132 versions (v1.0.1 to v1.8.1) over 107 days, averaging less than one release a day; from September to November it merged 232 PRs (Trendshift).

### What really sets it apart from others: a backend that lets you run a paid site

On creative features, Yingce is pretty much the same as its peers. Where it pulls ahead is the backend: user management, credits, redemption codes, top-ups via WeChat / Alipay / Hupijiao / Yi Pay, per-model "cost price" and "sale price," time-based pricing, and revenue statistics. Many CHANGELOG entries are about refining billing, for example v1.8.1 added "reference-video billing," and v1.7.3.7 priced cache hits and cache writes separately. The license is MIT, so you can modify it and use it commercially without open-sourcing your changes.

In other words, it's source code you can directly use to launch an "AI short-drama site" and resell model compute. Two numbers back up this judgment:

- **High fork ratio.** 545 forks / 1,675 stars = 0.33. Comparable projects: Toonflow 0.18, Huobao Drama 0.19, infinite-canvas 0.24, LocalMiniDrama 0.23 (GitHub, 11-06). More people fork it than just bookmark it with a star, which suggests they plan to deploy it themselves.

- **Sponsors are almost all API sellers.** The README lists 10: soonstudio, Fluxion AI, panda AI and Yuandian Power are API resellers / model distributors, and Metaso features the MiniMax H3 video API. Fluxion's signup link carries the `campaign=yingce` referral parameter, and the demo site's banner reads "MiniMax H3 at official price, 2% off." Anyone launching a site has to buy compute, and these vendors sell it, so sponsoring this repo is a good deal for them. The author doesn't charge users; the money comes from sponsor slots. Sponsorship fees are not disclosed.

A caveat: the README itself says "Best suited for personal, local, or trusted-environment deployment; do not expose it as a multi-user public service without security hardening." Anyone who actually runs it as a site has to add security themselves.

### Who it's competing with

In 2026, "novel-to-AI-short-drama" is one of the most crowded directions on Chinese GitHub. Star counts as of 11-06:

| Project | Stars | Compared with Yingce |
|---|---|---|
| Toonflow | 20.8k, +874 in the last 7 days | A category leader with similar features (canvas, Agent, 3D director's console), AGPL license, more of a desktop standalone app |
| Huobao Drama huobao-drama | 17.0k | one sentence, fully automated video output, fastest to get started |
| basketikun/infinite-canvas | 9.7k | nearly the same blurb, but leans toward image creation and skips the drama pipeline; indexed earlier than Yingce |
| printfilm / dramaclaw / Jellyfish | 7.1k–7.4k | both gain stars faster than Yingce; dramaclaw +677 in the last 7 days |
| waoowaoo | 6.8k in 4 days (Tencent Cloud Developer Community) | pitched as beginner-friendly, with explosive early growth |
| ArcReel | 6.2k | novel-to-storyboard, export to a CapCut draft; feature sets overlap heavily |
| LocalMiniDrama, LumenX (Alibaba), tigerowo/infinite-canvas and others | 1.4k–2.4k | same tier as Yingce |
| Jimeng Smart Canvas, LiblibAI's LibTV, SenseTime Seko, ByteDance Xiaoyunque | Closed source | big-tech web products, pay-per-use |

Yingce has roughly a tenth of the top project's stars. Blogger 火山哥 10-14 listed the "six most practical open-source AI drama tools" and left it out; a Sohu roundup 10-26 ranked it behind Huobao and Toonflow.

The reasons it hasn't scaled are visible: it entered six months later (July vs January); the name open-ai-canvas is too generic, and searches return OpenAI Canvas and langchain's open-canvas; deployment on Go + Bun + PG + Redis is heavy, so beginner lists rank it low; and it has no one-line hook. Huobao has "one sentence to a video," waoowaoo has "beginner-friendly," while Yingce only has "full features."

### If it were you

**C: worth studying, but building another AI drama workbench to chase stars isn't worth it.** The niche already has two leaders over 10k stars and a dozen projects in the thousands. Even with Yingce's effort (a release a day, five contributors), it sits around 15th.

**Good fit for** : people who want to run an open-source project and earn something from it. They can copy the "open source + admin back office + sponsor slots" model. Those already building open-source AI video or image tools can add the billing back office.

**Bad fit for** : anyone who wants to launch "yet another novel-to-drama workbench" to chase stars, anyone who can't keep up a high release cadence, and anyone hoping to beat the leaders on feature breadth.

**How to do it** : don't compete on features. Three things are worth copying. First, pick a generation niche that isn't crowded yet (AI e-commerce video, AI audiobooks, AI comics) and build an MIT version with a built-in credits and payments back office that can go live immediately, putting "operable" on the first screen of the README. Second, run a public demo site with test accounts, and keep a banner slot at the top. Third, open sponsor slots and approach API relay sites and model resellers directly, giving them links with referral parameters. Name it something searchable, not a generic open-xxx-canvas.

**How to get users** : follow its playbook. Post on X the day the repo goes up, with topic tags and screenshots. Get Chinese repost accounts like Geek Lite and knowledge-sharing accounts to share it; a post with ~10k views brings in one or two hundred stars. Write a Chinese update note for every major release, to feed repost accounts and roundup articles. If you want trending, prepare before launch, because the stars need to concentrate in the first week.

**When to move** : while AI video model prices are still falling and relay sites are competing for customers who want to open their own sites. Once big tech offers "one-click drama site" as a managed service, this approach stops working.

**Signs you're wrong** : (1) Yingce's daily star gain drops to single digits next month and forks stop growing, which means "can launch a site" isn't attracting people who actually launch sites; (2) the sponsor list shrinks or switches to companies unrelated to APIs, meaning sponsor slots aren't earning money; (3) someone ships a narrower niche (say, AI comics only) with a full back office and passes Yingce's star count within a month. That would mean switching niches was the right move, and copying the drama workbench was the mistake.

### Data and sources

- GitHub repo and README: github.com/ddcat-ai/open-ai-canvas (stars and forks as of 11-06; daily stars from 10-10); CHANGELOG.

- Trendshift: trendshift.io/repositories/94298 (stars, forks, PRs from 9-07 to 11-06).

- X: @liangwenhao3 (7-22), @QingQ77 (7-30), @vintcessun (8-01), @knowledgefxg (10-29), @huoshan007 (10-14); view counts as displayed on the page.

- Coverage and roundups: z-ab.com (10-04), CSDN (10-13), Sohu (10-26), Tencent Cloud Developer Community.

- Star counts for similar projects: GitHub, 11-06.

[GitHub repo ↗](https://github.com/ddcat-ai/open-ai-canvas)

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

Competitors

| Name | Where | Launched | Scale | How it differs from this one |
|---|---|---|---|---|
| Toonflow (HBAI-Ltd/Toonflow-app) | Open source | 2026-01-29 | 20.8k stars, +7 in the last 874 days | Leader in the niche, similar features (canvas + Agent + 3D director stage), but AGPL license and desktop-first, single-user; Yingce is MIT, web-based multi-user, with a billing back office |
| Huobao Drama (chatfire-AI/huobao-drama) | Open source | 2026-01-05 | 17.0k stars | One-sentence fully automated output, fast to start; Yingce is heavier, aimed at professional workflows and operations |
| basketikun/infinite-canvas | Open source | 2026-05-18 | 9.7k stars | Also called an AI infinite-canvas workbench, but leans toward image creation with no drama pipeline |
| ArcReel | Open source | 2026-02-07 | 6.2k stars | Also novel-to-assets, storyboards, CapCut drafts, with heavily overlapping features; Yingce adds operations billing and plugin channels |
| dramaclaw / Jellyfish / printfilm / waoowaoo | Open source | 2026 | each 6.8k–7.4k stars | Projects in the same niche that are gaining faster, each with a one-line hook (general-purpose engine, beginner-friendly, etc.); Yingce has no comparable single hook |
| LocalMiniDrama / BigBanana AI Director / LumenX (Alibaba) / tigerowo/infinite-canvas | Open source | 2026 | 1.4k–2.4k stars | Same tier as Yingce; LocalMiniDrama is purely local, LumenX is Alibaba's; Yingce's strength is that it's operable |
| Jimeng Smart Canvas / LibTV / Seko / Xiaoyunque / TapNow | Web | 2025–2026 | Not verified | Closed SaaS with pay-per-use pricing; Yingce is an open-source alternative you can self-host and resell |

Where users come from

- **Confirmed** **Chinese AI repost accounts on X, first-day reposts** : 7-22 @liangwenhao3, 7-30 Geek Lite @QingQ77, 8-01 @vintcessun posted, matching roughly 800 stars over the first 47 days

- Inferred **X post drove a 10-30/31 spike** : 10-29 @knowledgefxg posted, next day +44, day after +77; timing matches

- **Confirmed** **Chinese blogs / SEO reposts** : z-ab.com 10-04, CSDN gitcode mirror review 10-13, Sohu roundup 10-26, Threads, navigation-site listings

- **Confirmed** **Niche heat spillover / roundups** : through 2026, many open-source AI drama roundups appeared, and Yingce was picked up in passing (Sohu, Zhihu, CSDN)

- Inferred **API relay sponsor ecosystem** : the README lists 10 sponsors, mostly API relay sites. Fluxion carries the campaign=yingce referral parameter, and the demo site shows a Metaso MiniMax H3 banner. No evidence found of the other side sending traffic back

## Why Yingce open-ai-canvas (ddcat-ai/open-ai-canvas) took off

> Channel: GitHub · research day 2026-11-06 · data card: +7 stars in the last 193 days, 1675 total stars / 545 forks, repo created 2026-07-22 (107 days ago)

### Bottom line

Yingce didn't "take off" so much as land as a **second-tier player** in a crowded niche (open-source AI drama / micro-drama workbenches). Its 1675 stars came from three things: first, very granular work and extremely fast releases (132 versions in 107 days); second, Chinese tech bloggers and X repost accounts that kept sharing it in scattered waves; third, it is essentially **AI drama SaaS source code you can launch and charge for right away** : MIT licensed, with built-in credits, payments, channel pricing, and a back office. Its fork-to-star ratio of 0.33 is clearly higher than comparable projects, suggesting many people fork it to deploy and run it themselves, not just to bookmark it. The real money is in sponsor slots: the README and demo site carry a row of ads from API relay sites and model resellers.

### 1. What it is

- An open-source AI film and drama creation workbench. After you upload a novel, the pipeline handles adaptation, episode splitting, character design, and storyboards, then generates video segment by segment and composes the final cut. It also includes an infinite canvas (image, video, text, audio, and 3D previz nodes), a canvas Agent, a skill library, plugin-based model channels, and timeline editing (with CapCut draft export). The next version is in internal testing for a 3D "world engine." Source: [README](https://github.com/ddcat-ai/open-ai-canvas).

- The stack is React + Bun on the front end and Go/Gin on the back end. It uses SQLite locally and PostgreSQL + Redis for deployment, with a one-click install script. Live demo at [yingce.tv](https://yingce.tv) (test/test123456).

- **Key point: it ships a complete operations back office** , including user management, credits operations, redemption codes, payment top-ups (WeChat Pay / Alipay / Hupijiao / Yi Pay plugins), per-model "cost / sell price," time-of-day token pricing, and P95 and revenue stats. Many CHANGELOG entries refine billing, such as v1.8.1 adding "reference video billing" and v1.7.3.7 splitting the sell price for cache hits and cache writes. Source: [CHANGELOG](https://github.com/ddcat-ai/open-ai-canvas/blob/main/CHANGELOG.md). The CSDN review also lists "operable: built-in users, credits, payments, back office" as a selling point ([CSDN 2026-10-13](https://blog.csdn.net/gitblog_00095/article/details/167789590)).

### 2. Growth data

- In this database's github.repo_day: 2026-10-10 had 1161 stars / 367 forks, rising to 1675 / 545 by 11-06. Daily gains mostly ran 10–25; 10-30 +44 and 10-31 +77 (the single-day high in the observation window), then back to 16–25 per day.

- [Trendshift](https://trendshift.io/repositories/94298): from 9-07 to 11-06, +874 stars, +333 forks, and 232 PRs merged, with daily gains mostly 5–18; **it never entered Trendshift's top 25** . Working backward, the first 47 days after the repo was created (7-22 to 9-07) collected about 800 stars. That came from one wave after launch, matching several reposts on X from late July to early August.

- The repo was only 107 days old, short of the 60-day threshold, and it also didn't qualify for trending (trending_days_any=0), so it never got GitHub trending traffic.

- Release pace: the CHANGELOG records 132 versions (v1.0.1 to v1.8.1), averaging less than one release a day; Trendshift shows 5 contributors.

- Fork/star ratio = 0.33 (545/1675). Comparison: Toonflow 0.18, Huobao Drama 0.19, basketikun/infinite-canvas 0.24, LocalMiniDrama 0.23 (from this database's github.repo_profile, 2026-11-06).

### 3. Competitors (the landscape first)

Open-source AI drama and micro-drama workbenches are among the most crowded niches on Chinese GitHub in 2026. The star counts below come from this database's github.repo_profile (2026-11-06):

| Project | Form | Launched | Scale | How it differs from Yingce |
|---|---|---|---|---|
| HBAI-Ltd/Toonflow-app | Open source | 2026-01-29 | 20.8k stars, +7 in the last 874 days | Leader in the niche; also an infinite canvas, three-layer Agent, and 3D director stage; AGPL-3.0 license (commercial use requires open-sourcing), desktop-first |
| chatfire-AI/huobao-drama Huobao Drama | Open source | 2026-01-05 | 17.0k stars | One sentence to a fully automated video, fastest to start; no canvas or back office as heavy as Yingce's |
| basketikun/infinite-canvas | Open source | 2026-05-18 | 9.7k stars | Nearly the same blurb ("open-source infinite canvas workbench for AI creation"), but leans toward image creation with no drama pipeline |
| yi1108/printfilm | Open source | 2026-09-10 | 7.4k stars | AI short-video marketing + short drama, single-day peak +441 |
| dramaclaw/dramaclaw | Open source | 2026-03-27 | 7.3k stars, +7 in the last 677 days | General-purpose AIGC video engine; one script-to-video pipeline |
| Forget-C/Jellyfish | Open source | 2026-03-06 | 7.1k stars | End-to-end AI drama production workspace |
| ArcReel/ArcReel | Open source | 2026-02-07 | 6.2k stars | Agent-driven: novel to assets, storyboards, CapCut drafts, with features that heavily overlap Yingce |
| waoowaoo | Open source | 2026 | 4 days 6.8k stars ([Tencent Cloud](https://cloud.tencent.com/developer/article/2639493)) | pitched as beginner-friendly, with explosive early growth |
| LocalMiniDrama / BigBanana / LumenX (Alibaba) / AIComicBuilder / ai-fusion-video, etc. | Open source | 2026 | 1.4k–2.4k stars | Same tier as Yingce, each with a different focus (fully local, Alibaba ecosystem, Java) |
| tigerowo/infinite-canvas | Open source | 2026-06-30 | 1.5k stars | Infinite canvas + director console + panorama view, the closest in form to Yingce |
| Jimeng smart canvas, LiblibAI's LibTV, SenseTime Seko, ByteDance Xiaoyunque, TapNow, etc. | Web / big-tech features | 2025–2026 | Scale unverified | Closed-source SaaS with pay-per-use pricing; Yingce is a self-hostable version you can modify and resell |

Verdict: **Yingce ranks around the teens in its category** , with about 1/10 stars, roughly a fraction of the leaders. It's not a category definer. X blogger Huoshan Ge's 10-14 list of "the six most practical open-source AI short-drama tools" doesn't include it ([X](https://x.com/huoshan007/status/2110191359818596527)). Sohu's 10-26 roundup puts it behind Huobao and Toonflow, positioning it as "a more comprehensive film and TV production management software" ([Sohu](https://www.sohu.com/a/1092765781_121885030)).

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

| Channel | Evidence | Confirmed |
|---|---|---|
| Chinese-language AI repost accounts / tech accounts retweeting the launch | On the day the database was built (7-22), @liangwenhao3 posted ([X](https://x.com/liangwenhao3/status/2079904076410819047)); 7-30 Geek Lite @QingQ77 ([X](https://x.com/QingQ77/status/2082806916615688685)); 8-01 @vintcessun ([X](https://x.com/vintcessun/status/2083428780752928804)); and there's also a post from Felix (俊哥). This lines up with the early ramp: about 800 stars in the first 47 days | Yes |
| 10-29 X post drove a small spike of 10-30/31 | @knowledgefxg ("Knowledge Sharing Officer") posted about Yingce on 10-29 ([X](https://x.com/knowledgefxg/status/2115785822792507470)); the next day +44, and the day after that +77. The timing matches, but causation is unverified. | No |
| Chinese blogs / SEO reposts | z-ab.com's recommendation article 10-04 (at the time 1.1k stars) ([link](https://z-ab.com/2026/10/04/yingce-open-ai-canvas/)); a CSDN review via a gitcode mirror (10-13); a Sohu roundup 10-26; a Threads repost (950 stars at the time); included by navigation sites like Xiami Partners and tg.noisework. | Yes |
| Spillover from overall category buzz | AI short dramas / animated dramas have been huge through 2026, and video models like Seedance 2.0 launched, driving a flood of "open-source short-drama tool roundup" content; Yingce got picked up incidentally in all kinds of roundups. | Yes (at the category level) |
| Sponsor / API relay station ecosystem cross-promotion | The README lists 10 sponsors, mostly API relays / model resellers (soonstudio, Fluxion AI with an invite parameter `campaign=yingce`, panda AI, Yuandian Dongli, and MiniMax H3 distribution from Metaso). The demo site's header carries a banner for MiniMax H3 at "official price 2% off". These vendors have an incentive to push Yingce to their own users, but I found no direct evidence of them actively promoting it. | No |
| Skill ecosystem cross-links | The README recommends the community skill pack itsWyatt-K/judian-skills; short-drama skill libraries (shuohao-skills 6k stars, drama-skills 3.8k stars) are very popular, and Yingce offers an entry point to "install skills from GitHub". | Yes (cross-links exist; traffic referral not verified) |
| GitHub trending | None. It couldn't chart within the first 60 days and hasn't since. | — |

No evidence found of paid promotion, HN, Product Hunt, or dedicated Bilibili / Douyin videos (search coverage is limited).

### 5. Why it reached 1.7k (rather than more or fewer)

1. **Rode a hot category** : "novel-to-AI-short-drama" has been one of the hottest directions in Chinese open source through 2026. The leaders Toonflow and Huobao both exceed 17k stars, and the whole category keeps getting roundups and reposts. Yingce rode the tailwind.

2. **Heavy, detailed product** : 107 days, 132 versions, 5 contributors, and 232 PRs merged in two months. It covers scripts, settings, storyboards, per-segment video, timelines, 3D previsualization, agents, and skills, with professional-looking README screenshots. One blogger was willing to write that it's "real film-and-TV workflow, not an image-generation toy" (z-ab).

3. **Operable** : this is the biggest difference from most peers. MIT license plus a billing / payment / channel pricing / redemption code backend means it's effectively source code you can use to launch a paid "AI short-drama site" right away. The high fork ratio (0.33) and the sponsor makeup (all API sellers) both suggest the real users are relay stations and small studios that want to self-host sites and resell compute, not just creators.

4. **Why it isn't bigger** : it started about half a year after the leaders (July vs January); the name open-ai-canvas is too generic and easily confused with langchain-ai/open-canvas and OpenAI Canvas, so it gets buried in search; it never made trending; there was no single viral X post from a big account; and the heavy deployment (Go + Bun + PG/Redis) is unfriendly to beginners, so beginner recommendation lists rank it lower.

### 6. Differentiation

- Compared with Toonflow / Huobao / ArcReel, **the creative features are broadly similar** : all have an infinite canvas, storyboarding, character consistency, and a 3D director console.

- What's actually different: **(a) MIT license plus a full commercial operations backend** (credits, payments, per-model cost and sale pricing, time-of-day pricing, Volcano Engine real-person asset verification), positioned as "source code for running your own site"; **(b) a plugin-based model channel system** (declarative plugins, 100+ official plugin packs, adapted for relay stations); **(c) a web-based, multi-user, server-side async task model** , not a desktop standalone tool.

- **This differentiation explains its fork ratio and sponsorship revenue, but not its star count** . Stars came mainly from category heat and repost accounts. The differentiation let it survive in a crowded category and attract API vendor sponsors, but it didn't push it to the top.

### 7. Visibility and copycats

- Visibility: **circulating in small circles** . Only Chinese X repost accounts, personal blogs, and CSDN / Sohu auto-generated articles have written about it. No big account or media outlet has covered it specifically, and in the leading roundups it sits in the "also worth noting" tier. No public revenue figures, and the dev team (ddcat-ai, an org with no public members) hasn't surfaced.

- Copycat dynamics: similar-concept projects already existed in batches (see the competitor table), and **it is itself one of the copycats** . The descriptions, look, and form of basketikun/infinite-canvas and tigerowo/infinite-canvas are very similar to it, and the order of precedence is unclear (basketikun was created earlier). Forks include personal ones like daoge05178; whether any were modified and launched commercially is unverified.

### 8. What others can learn

- The combination of **open source + operations backend + sponsor slots** : it makes "can run a site and earn money" the selling point, attracting relay stations / API resellers to sponsor and embed invite links (Fluxion's `campaign=yingce`). The author doesn't collect user money, monetizing through B2B ad slots instead.

- A public demo site plus a test account lowers the barrier to trying it, and the banner at the top of the demo site is itself an ad slot.

- High-frequency releases plus a detailed Chinese CHANGELOG keep feeding repost accounts with "just updated" material.

- Lesson learned: the repo name needs to be distinctive and easy to search; entering a crowded category late, feature stacking alone won't reach the top. You need a hook that can be summed up in one sentence (like waoowaoo's "beginner-friendly" or Huobao's "one sentence to a finished video").

### 9. Uncertain

- Whether the 10-31 spike of +77 was actually caused by @knowledgefxg's post (only the timing matches).

- How much sponsorship revenue there is, and whether sponsors channel traffic back.

- How many forks have actually been deployed as paid public sites.

- Team identity (who is behind ddcat-ai, and what the relationship is to sponsors like "Happy Machine Art Lab").

- The GitHub API was rate-limited that day, so the stargazer breakdown, contributor list, and release times weren't checked live. Daily star curves for the early period come from Trendshift and back-calculation, not direct data.

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