Stringers
jaikydota/kk-mix · jaikydota

365 stars and not one person has written about it: how a Windows video-mashup tool "grew"

What it is A Windows desktop program that uses FFmpeg to run 15 kinds of batch edits, each taking a folder in and writing a folder out. Its main pitch is interleaving several batches of footage into dozens of distinct short videos, aimed at account farms.

Why it took off It didn't, really. A marketing-tools company open-sourced a mashup app it had been selling as license keys. Since then, stars have trickled in at a steady rate, with an occasional single-day drop of 16. The forks and issues come from batches of freshly registered accounts, and no one has published a write-up of it anywhere online.

DHow you could follow it Don't copy it as-is. If you really want to build in this category, fill the gaps it leaves: a packaged exe, dedup that passes platform originality checks, direct export to CapCut drafts, and customer acquisition through screen recordings on Bilibili, Douyin, and operator groups.

Grade D: its growth is manufactured, not a demand signal. The category already has mature open-source projects with 7k to 140k stars, and the commercial side has a row of paid software and big-platform APIs. Copying it won't get you real stars. Few people know about it yet
Data (updated daily) · Repo created 2026-09-11 · 526 stars · 7 days +110
GitHub · stars
020040060010-0910-2211-0111-1011-20526
Data 2026-10-09 → 2026-11-20 (43 days, 37 observations). Points before 10-28 come from deeply / archive backfill, not every day has one

kk-mix is a batch video-mashup tool for Windows. It picked up 365 stars in two months, and the pace has been accelerating over the last week. It looks like an unreported project that has spread through small circles on its own. Going through its issues, forks, and daily star counts gives the opposite answer: most of those stars were bought. This piece covers what it is, where the stars came from, where the real demand in this category sits, and what someone chasing stars with a similar project should copy and what they should avoid.

Three questions from early on Oct 8

On Oct 8 03:27 (UTC) in 2026, a new GitHub account called bth5Joy was registered with the bio "nothing to say". 33 minutes later it opened a second issue under kk-mix: "Does it support batch image resizing?" Half an hour earlier, another account, addichen014, had filed the first issue asking how to deploy a TTS voiceover service. That afternoon a third account, chb0222, asked why batch compression made images blurry. It was also registered that same day.

All three accounts have only two repos: a fork of kk-mix, and a fork of BullFlow, another project from the same author. All four issues are in English and follow one template, opening with "Hi, I'm trying to use…" and then listing 1, 2, or 3 questions. The author replies in a line or two and closes each one.

kk-mix was created in Sep 11. For almost a month before that, it barely moved. From Oct 8 on, the first fork appeared and stars started climbing.

What it is

A desktop app built with Python and PySide6 for the UI and FFmpeg doing the work. It offers 15 batch operations, all following the same pattern: pick an input folder, get an output folder.

  • Interleaved splicing: given N folders, the i-th clip from each folder is combined into the i-th output video. The README example is 20 opening clips, 20 product shots, and 20 closing clips. That produces 20 distinct finished videos in one run, with 20 transition styles.
  • Voiceover splicing: each folder gets a script, one line per video. Lines are read aloud in the same cloned voice using a self-hosted Index-TTS, and clips are automatically trimmed to match the voiceover length.
  • Split screen, picture-in-picture, batch segmentation, speed changes, rotation, cropping to 9:16, titles, watermarks, background music, volume adjustment, compression, transcoding, and frame extraction.

The README's "who uses it" table lists "short-video matrices / multi-account distribution" in the first row. The interface is bilingual (Chinese and English), styled in Fluent, with full screenshots.

But the repo has no releases and no packaged exe (GitHub Releases shows 0, checked live 11-09). To use it you have to install Python yourself, run uv sync, then download ffmpeg and ffprobe and drop them into the folder. Most of the e-commerce sellers and account operators it's aimed at won't get past that step.

Where it came from

Looking at the commit history, this was not an open-source project written from scratch. Starting in April 2026, it was private commercial software: on Apr 14, "changed to license-key logic"; on 15 April, "added one-license-per-machine mechanism" and "added pyd decompilation barrier". In late April it was renamed "Super Large Editing Tool", and in May "Super Large Mashup System". It was a mashup app sold as machine-locked license keys.

Within a single day in Sep 11, the author pushed 11 commits: "open-source prep: clean sensitive info", 364 bilingual translation commits, a README rewritten with graphics, and a screenshot script, all co-authored with Claude. After that, no updates. The license-key code is still in the repo, just disabled by default at build time.

The author, jaikydota, has had the account since 2014. In 2016 they published a set of Android widget libraries, one of which has 210 stars earned organically. The public profile lists a vice director at a Beijing internet company. On Oct 14, the same account open-sourced BullFlow: enterprise knowledge base, marketing content production, and distribution to 17 platforms. That corresponds to a commercial marketing tool with a price sheet and a Telegram sales contact. Both projects share a style: long bilingual READMEs written with AI help, pushed all at once on release day and then left alone. It looks like a marketing-tools company open-sourcing its internal products one by one to build GitHub reputation (inference).

How the stars grew

Datekk-mix starChangesBullFlow stars over the same period
10-0920Just getting startedNot yet created
10-16119+99 in a week—
10-17 ~ 10-22142 → 232+16 ~ 23 per day21 → 86
10-24216Day -16106
10-25 ~ 10-31217 → 2360 ~ 6 per day118 → 237
11-01 ~ 11-08239 → 343+3 ~ 18 per day257 → 400 or so
11-09 (checked live)365428

(The daily star counts were recorded by us day by day; 11-09 was checked live on GitHub.)

Three things look off:

  1. No spike. A project that spreads organically usually follows a curve where it gets reshared one day, jumps to dozens or hundreds, then falls back. kk-mix's biggest day was Oct 17 +23, and the other days hover around a dozen, like a tap dripping at a steady rate.
  2. Oct 24 lost 16. Real people rarely remove stars in batches. When GitHub periodically cleans out fake accounts, the stars they gave disappear all at once, which is the most common trace of star-buying.
  3. The two repos move in sync. BullFlow was only created on Oct 14, yet it climbs at almost the same steady pace as kk-mix, rising and slowing together.

GitHub no longer publishes the list of people who starred a repo, but the fork list is still visible. Roughly 20 of the 25 fork accounts look like batch accounts at a glance: their creation dates cluster on the same days (2025-08-09 has 4, 2026-01-23 has 3, 2026-04-06 has 3), and nearly all have 0 followers. The few I checked had nothing but forks, and they repeatedly forked the same unrelated repos (a CRM, a model collection, a Solidity audit tool) with no connection to video. Their bios are template fake personas like "weekday software engineer, weekend hardware hobbyist | ESP32". There's also a star-selling site that generated a page for kk-mix with plans like "+50 stars from $20, delivered over 3 days". Such sites auto-generate pages for many repos, so that alone doesn't prove the author bought stars. It only shows that this kind of steady-drip service is readily available to buy.

Searching further: in web search, kk-mix shows up only on a few auto-generated "new star" leaderboard sites built from GitHub data. No article about it on Juejin, Zhihu, CSDN, Bilibili, or WeChat Official Accounts. On X, searches for "kk-mix" and "jaikydota" over the last 60 days return 0 results each. Of the 25 forks, only one looks like a real user: someone with thirty-plus video and scraping tools in their account (MoneyPrinterTurbo, NarratoAI, CapCut draft automation…) who is systematically collecting tools in this space.

So the answer to "why did it take off" is simple: it didn't. The stars were bought, the issues were filed by the author's own people, and the number of people who have actually seen it is probably single digits (inference).

The demand is real, but the space is already crowded

Setting aside the star question, the demand kk-mix targets is legitimate. Teams running short-video account matrices and Qianchuan ad campaigns in China need to break a batch of footage apart and recombine it into dozens or hundreds of "different" videos. That's an established paid market. The problem is that the path is already packed:

NameWhereScaleHow kk-mix differs
suifengqjn/videoWaterOpen source, 20191,187 starAlmost identical: batch transcoding, subtitles, watermarks, cropping, speed changes, segmenting, compositing, BGM. Older libraries that built up stars organically
ddean2009/MoneyPrinterPlusOpen source, 20247,324 starAI batch generation + mashups + automatic publishing to Douyin, Kuaishou, Xiaohongshu, and WeChat Channels. Has real word-of-mouth on Bilibili and Zhihu
harry0703/MoneyPrinterTurboOpen source, 2024About 140k starsTakes a topic and generates short videos from scratch with AI. kk-mix processes existing footage
GuanYixuan/pyJianYingDraftOpen source, 20244,850 starBuilds CapCut drafts with Python for a mashup pipeline, aimed at people who code
A batch of projects including JJYB_AI_VideoAutoCutOpen source, 2025–202680 ~ 1,200 starMashups, dedup, and Qianchuan creative material, often with AI narration or dedup algorithms
Super Director, Kuaijing, Huaer Duoduo, Hunting Bee Mashup, MIXCUTCommercial software / SaaSA mature paid market, with annual fees reportedly starting at several thousand yuan (unverified)Includes "dedup" to pass platform originality checks, account-matrix management, and batch publishing
CapCut marketing-video export, Alibaba Cloud and Volcano Engine mashup APIsbig techBig-company tierCloud-based template batch mashups, marketed directly around matrix distribution

The differences kk-mix can actually name are: form-filling instead of timelines, 15 batch operations in one bilingual interface, free local use, and an MIT license. But it lacks the part where commercial software actually makes money: dedup, passing originality checks, and batch publishing. Open-source peers in the category already include AI copywriting and automatic publishing. kk-mix is a subset of this category, and it's the subset you have to set up yourself.

What to copy and what not to

What to learn: the low-cost way to open-source an internal tool. Strip out the licensing, have AI write the bilingual README in one day, include a scenario table, a hands-on example ("3 folders in, 20 videos out"), and a screenshot script, then push it and stop maintaining it. kk-mix's README structure is genuinely good: it opens with "the timeline editor is for one video, this is for a different situation", then a "who uses it" scenario table, then features grouped by purpose, and finally an example you can run by following along. That structure is worth copying.

What not to learn: its stars. Bought stars don't bring real issues, PRs, or reshares, and they won't get a project onto trending. kk-mix has been up for 60 days and has never once appeared on trending. A single GitHub cleanup knocks a chunk of them off. For an individual developer trying to build reputation, these are the least useful stars there are.

Bonus tip: how to tell if a repo's stars are fake. When the stargazer list isn't visible, these checks still work: do the fork accounts' creation dates cluster in batches, are they all forks, and do they fork the same unrelated repos; do issue authors open their first issue the day their account was created; does the star curve show spikes or a sudden one-day drop; can you find any human-written write-up of the repo online.

If it were you

Verdict: D. Worth a look, not worth copying. kk-mix's "breakout" isn't a demand signal, it's bought numbers. Its category already has mature open-source projects with 7k to 140k stars, and on the commercial side there's a whole row of paid software and big-tech APIs. Building a clone of the same FFmpeg batch tool won't get you real stars.

Who it suits: developers on short-video matrix or ad-creative teams with real batch-editing needs and frontline feedback, who can build the features commercial software has.

Who it doesn't suit: anyone planning another "FFmpeg batch + nice UI" tool and waiting for stars; anyone without a Chinese community channel to post in; anyone planning to start by buying stars.

If you still want to build in this space, go after what it does badly:

  1. Package it. A single exe that runs on download, with FFmpeg bundled, no Python needed. kk-mix's README says it "can be packaged into a single folder" but never ships a release. That's exactly where target users get stuck first.
  2. Build dedup. What commercial competitors actually charge for is getting recombined videos past platform originality checks (combinations of frame swaps, mirroring, speed changes, noise, BGM swaps). Few open-source projects do this well or explain how it works.
  3. Connect to CapCut (Jianying). Don't build your own timeline; generate CapCut drafts directly so operators can finish in software they already know. pyJianYingDraft has already shown developers are buying in. What's missing is the layer for people who don't write code.

How to get users: this category's real users are on Bilibili, Douyin, Xiaohongshu, and e-commerce operator groups, not GitHub. A screen recording of "3 folders, 20 unique videos in one minute" will do more than a README. Answer specific questions where projects like MoneyPrinterPlus get discussed.

When to move: no rush. This need has existed for years and won't disappear because of kk-mix. But big-tech cloud remixing keeps cutting prices, so the "free, runs locally" pitch will gradually weaken.

What would prove this read wrong: (1) human-written write-ups, human-filed issues or PRs show up for kk-mix, and its star curve gets a spike that tracks a specific post, meaning real users are seeing it and the source call needs revising; (2) someone ships a packaged, open-source remixing tool with dedup that reaches 2,000 stars in three months from a couple of screen recordings, meaning this lane is worth more than a D and the grade goes up.

Data and sources

GitHub: jaikydota/kk-mix repo page, README, commit history (the license-code and rename commits starting 2026-04, the 11 commits on open-source day 09-11), Releases (0), registration dates of the 4 issue authors and askers, public profiles of the 25 fork accounts (checked 11-09 to 11-10); jaikydota/BullFlow (11-09 star count checked 428). Daily star counts are our own daily logs. Star counts for comparable repos were checked in November. Author identity comes from their public profile. Web and X search as of 11-09. Anything marked "inferred" or "unverified" is an inference or unchecked.

GitHub repo ↗

Research notes: competitor table, acquisition evidence, full research report
Competitors
NameWhereLaunchedScaleHow it differs from this one
suifengqjn/videoWaterOpen source2019-111,187 stars / 222 forksNearly identical video batch-processing tool; kk-mix adds interleaved multi-folder stitching to produce N distinct videos and TTS voiceover splicing, with an updated UI.
ddean2009/MoneyPrinterPlusOpen source2024-057,324 stars / 1,254 forksAI batch generation + remixing + automatic multi-platform publishing, broad feature set with real traction. kk-mix only does FFmpeg batch processing, with no AI copywriting or publishing.
harry0703/MoneyPrinterTurboOpen source2024About 140k starsTopic → AI generates short videos from scratch. kk-mix batch-processes existing footage.
GuanYixuan/pyJianYingDraftOpen source2024-074,850 starsGenerates CapCut drafts with Python for a remixing pipeline, aimed at developers. kk-mix targets people who don't write code.
JJYB_AI_VideoAutoCut / VideoMatrix-- / mixcut-windowsOpen source2025-11 ~ 2026-051,214 / 87 / 82 starsComparable remixing / dedup / Qianchuan creative tools, mostly with AI narration or dedup. kk-mix has no dedup and no AI storyboarding.
Chaojiyoubian / Kuaijing / Huaerduoduo / Hunjianbee / MIXCUTOtherYearsMature paid SaaS and Windows software in China (recall card says annual fees start at ¥8880, unverified)Includes platform dedup algorithms, matrix account management, and batch publishing. kk-mix is free and local, and its features are a subset.
CapCut marketing video / Alibaba Cloud IMS remixing / Volcano Engine KickartBig-company featuresLiveBig-company tierTemplate-based batch remixing in the cloud, explicitly serving matrix-account publishing. kk-mix is an offline desktop version.
Where users come from
  • Confirmed Paid fork/star inflation: of the 25 fork accounts, about 20 are batch accounts: creation dates cluster (2025-08-09 four, 2026-01-23 three, 2026-04-06 three, etc.), most have 0 followers, all of them are forks that fork the same batch of unrelated repos (Wukong-AICRM, AgnesAI-Models, VexDB-Lite…), bios are templated. The star curve is linear with no spikes, and 10-24 it dropped 16 stars. The star-selling site sg-tree.eu has a page for this repo (supporting evidence only).
  • Verified self-made issues: 4 English issues all came from addichen014, bth5Joy, and chb0222; the latter two 2026-10-08 asked questions on the day they registered, and their only forks were of the author's kk-mix and BullFlow.
  • Confirmed Same-author repos vouching for each other: BullFlow (the open-source version of Chuhaibang Tuiliangniu) was open-sourced at the same time, uses the same front accounts, and gains stars at the same pace.
  • Inferred Content / community spread (X, Zhihu, Bilibili, Juejin, etc.): Jina, Exa, and Brave searches turned up no human write-ups. x-radar shows 60 posts in 0 days. Only one possibly real user, zhaoliuwanrong, has collected similar video tools systematically.

jaikydota/kk-mix: why it got stars

Research date 2026-11-09 (task baseline date 2026-11-07). Channel: GitHub. Data card: cards/gh_jaikydota__kk-mix.json.

Bottom line first

kk-mix is an open-source release of a Windows remixing tool that a company used to sell as license-key software (once called "Chaodabei Editing Tool / Chaodabei Remix System"). The license was stripped out and a bilingual README was written with AI. Its stars largely did not grow organically:

  • 365 stars (GitHub API, 2026-11-09), yet no human-written write-up anywhere online, no X / Xiaohongshu / Bilibili / Zhihu reposts, only automated aggregator sites and a site that sells stars generating pages for it.
  • The star curve is a steady drip, not a spike: after 10-16 it climbs a flat +15~23 per day, rising in sync with BullFlow, a new repo from the same author 10-14; 10-24 stars dropped from 232 to 216 (-16), the classic trace of GitHub purging fake accounts;
  • Of the 25 fork accounts visible, about 20 are batch accounts (0 followers, creation dates clustered, all forks of the same set of repos). All 4 issues are in English and were filed by accounts created the day they asked, whose only forks are the author's two repos.

So the answer to "why did it take off" is inflation (paid stars/forks plus self-staged issues), not content spread and not a product people discovered. The product itself targets a real need (short-video matrix remixing is a mature paid market in China with plenty of SaaS), but there's no evidence yet that real users are seeing this repo. The recall card's claim of "niche Chinese-community organic spread" doesn't hold up.

What it is

  • A Windows desktop tool built on Python + PySide6 (Fluent style) + FFmpeg, with 15 kinds of "folder in, folder out" batch processing: transition stitching (for N folders, file i from each folder gets combined into one video, 5 folders × 20 clips = 20 distinct videos), voiceover splicing (self-hosted Index-TTS voice cloning), split screen, picture-in-picture, and batch segment cutting, speed changes, rotation, aspect-ratio crops, titles, watermarks, BGM, volume, compression, transcoding, and frame extraction. The README's first scenario is "short-video matrix / multi-account distribution." (README, github.com/jaikydota/kk-mix)
  • No Releases and no packaged exe. Users have to run uv sync and download ffmpeg/ffprobe themselves. That's a real barrier for its target users (e-commerce and matrix-account operators).
  • The licensing system (per-machine binding to motherboard UUID + CPU ID, AES license codes, keygen.exe) is still in the repo, just disabled by default at compile time.

Origin: internal commercial software → open-sourced 9-11 within a day

Commit history (GitHub API commits):

  • Starting 2026-04-09 it was a private project with commits like "add batch cropping", "fix login bug", "switch to license-code logic, comment out online login" (04-14), "add per-machine binding" (04-15), "add pyd decompile barrier" (04-15), "rename to: Chaodabei Editing Tool" (04-27), "rename to: Chaodabei Remix System" (05-06), "remove unused llm and tts calls" (07-29). This was commercial remixing software sold with license codes.
  • On 2026-09-11 it pushed 11 commits in one day: "open-source prep: scrub sensitive info… add README/LICENSE", bilingual i18n (364 translated entries), an image-based README and screenshot script… each commit carries Co-Authored-By: Claude Opus 5 / Claude Fable 5.1. No code has been pushed since (pushed_at 2026-09-11 09:55).

The author jaikydota (Jaiky / Ou Jie) registered the account 2014 years ago, and in 2016 shipped a batch of Android widget libraries (Android-ImagesPickers, 210 stars, a real track record). The Facebook profile lists "Deputy Director, Beijing Hetian Yuxiang Internet Technology Co., Ltd" (facebook.com/jaikydota). The same author open-sourced BullFlow (推量牛: enterprise RAG, marketing content production, 17 platform distribution, 425 stars) in 2026-10-14. The commercial product behind it is 出海帮 · 推量牛 (ai365-web.chuhaibang.com), with a pricing plan and Telegram contact. kk-mix and BullFlow were open-sourced one after the other with the same style (AI-written bilingual long READMEs, promoted hard the day they went up, then never updated). It looks like a marketing-tool company batch-open-sourcing internal products to buy attention.

Growth data

Source: our github repo_day table (data migrated from deeply for 10-09 onward, daily snapshots from 10-28), plus current GitHub API values.

Datekk-mix starsDaily gainBullFlow stars
10-0920–(not created)
10-1135+11
10-161195 days +84
10-17 ~ 10-22142→232Daily +16~2321→86
10-24216-16106
10-25 ~ 10-31217→2360–6 per day118→237
11-01 ~ 11-10239→359Daily +3~18257→425
11-09(API)365
  • From creation on 9-11 to 10-08 there was almost no activity. On 10-08 the first issue and first fork appeared, and stars started rising from there. The starting point was a deliberate launch move, not someone's reshare.
  • No single-day spike (largest day +23,10-17), the typical shape of staggered paid delivery. On 10-24 it dropped 16 stars in one go, the usual pattern when GitHub deletes fake accounts.
  • Over the last 7 days +101 versus +39 in the prior 7 days, an "acceleration"; over the same period BullFlow was also rising steadily at +15–30/day. The two repos follow the same rhythm.
  • Never hit trending (created 60 days ago, just reached the threshold; this has no meaning by itself).

Growth method: evidence

1. Paid fork / star inflation (high confidence)

The stargazer list is currently not publicly available from GitHub (REST 404, GraphQL returns empty), so only forks can be checked. 25 fork accounts (GitHub API, 2026-11-10):

  • Creation dates cluster in batches: 2025-08-09 share 4 accounts (notomarom, of-ahenel, Luna7hadow, frashfiejc), 2026-01-23 share 3 (devmikets, calanipira19, nothimozglupp), 2026-04-06 share 3 (zhuyg2tw, ppc24, maniminun), 2025-04-15 share 2, 2026-01-16 share 2, 2026-05-04 share 2, 2026-10-08 share 2.
  • Most of the 0 fans have 6~34 repos on their account. The ones I checked, notomarom, of-ahenel and Luna7hadow, are all forks, and they fork the same batch of unrelated repos (Wukong-AICRM, AgnesAI-Models, VexDB-Lite, solidity-cot-auditor, HanyuanWang/LiveStream-Agent-Studio…). None of it has anything to do with kk-mix's video mixing. This is a star-farming service's client list.
  • Bios are templated fake identities: "Software engineer by day, hardware tinkerer by night | IoT enthusiast | ESP32 projects", "Technical writer | API documentation", "Beijing No 8 High School".
  • Fork pace is as steady as the star pace: 11-01 to 11-10, about 1~3 per day.
  • Side evidence: the star-selling site sg-tree.eu has a audit/jaikydota/kk-mix page (packages from +50 stars for $20, with a "3-day feed schedule" that drips them out by day). Sites like this auto-generate pages for many repos, so this alone doesn't prove the author bought stars. It only shows that this kind of steady-drip service is readily available.

2. Self-staged issues (high confidence)

4 issues, all in English, all in that LLM-style "I want to use feature X, can you give: 1. … 2. … 3. …" tone:

  • #1, #4 addichen014 (2023 signup, 3 repos: one awesome-list fork, a BullFlow fork, and a kk-mix fork);
  • #2 bth5Joy, #3 chb0222: both signed up 2026-10-08 and filed issues the same day. Their only forks are kk-mix and BullFlow, and their bios say "nothing to say" and "building something useless".
  • The author replies with a single line and that's it. These 3 accounts also forked BullFlow, which props up both repos with a "people are using this" facade.

3. Organic channels: basically none

  • Web search (Jina / Exa / Brave): kk-mix only appears on repositorystats, srclog, gitofanton (auto-generated "Hidden gems / Rising" lists), and sg-tree, sites that auto-generate from GitHub data. No one has written about it on Juejin, Zhihu, CSDN, Bilibili, WeChat Official Accounts, sspai, or HelloGitHub-type outlets.
  • X: across the x-radar corpus from the last 60 days, searches for kk-mix / jaikydota / BullFlow / 推量牛 returned 0 results in total.
  • The only fork that looks like a real user is zhaoliuwanrong. Their 30+ forks are all video and scraping tools (MoneyPrinterTurbo, NarratoAI, pyJianYingDraft, VideoMatrix, Spider_XHS…). This is someone systematically collecting tools in the same category, which shows the category does have real followers on GitHub, just a small number.
  • BullFlow has a LinkedIn post and an aiboss88 listing. kk-mix has neither.

Competitors

NameWhereLaunchedScaleHow it differs from kk-mix
suifengqjn/videoWaterOpen source2019-111,187 stars / 222 forksIt's almost the same thing: batch video processing (bitrate, transcoding, subtitles, watermarks, scrolling text, cropping, speed change, segmenting, compositing, BGM). It's an older repo that built up organically. kk-mix adds "interleaving multiple folders to generate N different videos" and TTS voiceover splicing, and has a newer UI.
ddean2009/MoneyPrinterPlusOpen source2024-057,324 stars / 1,254 forksAI-driven one-click batch generation + mixing + auto-publishing to Douyin, Kuaishou, Xiaohongshu, and WeChat Channels, with local TTS/ASR. Broader feature set and real traction (Bilibili, Zhihu, the author's blog flydean.com). kk-mix has no AI copywriting and no publishing; it's pure FFmpeg batch processing.
harry0703/MoneyPrinterTurboOpen source2024140K starsTopic → AI-generated short video. It positions itself as "generate from scratch", whereas kk-mix is "batch-process existing footage".
GuanYixuan/pyJianYingDraftOpen source2024-074,850 starsPython generates CapCut/JianYing drafts to automate mixing pipelines, aimed at developers. kk-mix targets people who don't write code.
JJYB_AI_VideoAutoCut / w2968066/VideoMatrix-- / RoshanGH/mixcut-windowsOpen source2025-11 / 2026-05 / 2026-051,214 / 87 / 82 starsSame family of "mixing / deduplication / Qianchuan material" tools, many with AI commentary and dedup algorithms. kk-mix has no dedup and no AI storyboarding.
Super Director, Kuaijing, Huaer Duoduo, the Mixcut Bee (Microsoft Store), MIXCUT, and othersCommercial SaaS / Windows softwareYearsA mature paid market in China. One recall card quotes annual fees starting at ¥8880 (unverified).All of them have a "dedup algorithm", matrix account management, and batch publishing, and sell to agencies and merchants. kk-mix is free, local, has no dedup and no publishing. Its features are a subset of theirs.
JianYing (marketing videos / batch templates), Alibaba Cloud IMS mixing, Volcano Engine Kickart video mixingBig-company features / APIsLiveBig-company tierCloud-based template batch mixing, specifically pitched for "matrix account distribution" scenarios. kk-mix is an offline desktop app.

Bottom line: this is a category that's crowded. Open source has mature projects with 7k~140k stars, commercial SaaS has a whole row of players, and big tech offers APIs. kk-mix has no unique capability.

Differentiation

  • The differences you can actually name: ① a minimal interaction of "fill in a form instead of a timeline, folder in, folder out"; ② one tool packs 15 kinds of FFmpeg batch processing together, bilingual, with a Fluent interface; ③ local and offline, MIT licensed, zero licensing fees (whereas commercial rivals charge annually).
  • These differences are not why it got its stars. No external discussion praises any of these points, and the star curve doesn't track any event. The real "differentiator" is that the author was willing to pay to inflate the stars.
  • Downsides: no packaged exe, so users have to install Python/uv/ffmpeg themselves. That's not friendly to non-technical users like matrix-account operators. It also lacks the core thing commercial rivals sell: dedup to get past platform originality checks.

What to take from it

  1. A cheap way to open-source an internal commercial tool: strip the licensing → have AI write a bilingual README, scenario tables, worked examples, and screenshot scripts in a day → push it out. The code no longer needs maintenance and still works as a facade. This README template (why it exists / scenario table / a worked example of 3 folders → 20 videos / output naming table) is genuinely well written and can be copied directly.
  2. Same company open-sources in batches and cross-promotes: kk-mix and BullFlow share an author and the same batch of self-staged accounts, used to build GitHub credibility for the company's products (Tuiliangniu / the overseas expansion helper).
  3. How to spot star-farming (more useful for our recall): check whether the fork accounts' creation dates cluster in batches and whether their repos are all forks of each other. Check whether issue authors filed questions on the day they signed up. Look for spikes or sudden drops in the star curve. Search the web for a human-written write-up. These checks still work even when the stargazer list isn't public.

Still unclear

  • We can't get the stargazer breakdown, so how many of the stars are real people can only be inferred from forks. Judging from forks, genuinely interested people number in the single digits.
  • Whether the author bought the stars themselves or the company's marketing team / an outsourced agency did it, we don't know. The sg-tree page only proves that this kind of service exists, not that a transaction happened.
  • Unverified: how well the "super-large-cup mixing system" sold before it was open-sourced, and whether the company's accounts on Douyin / Xiaohongshu / WeChat groups promoted kk-mix. Those platforms are outside our search capabilities (TikHub hasn't searched domestic Douyin).
  • The recall card says "the author has 19 repos, all repeating the same pattern". Most of the 19 repos are 2016-year-old Android repos. Only kk-mix and BullFlow share this pattern.

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