A Douyin account with 199 followers, one line "the mountain that boosts frame rate the most," and nearly 20,000 downloads in four days: Pavise's players are on Douyin, and the stars are just the overflow

Pavise is a free, open-source Windows "game resource manager" built by an individual developer in Shanghai, accepting only voluntary donations. Its stars rose from over five hundred to over eight hundred in the past week, but behind that gain are tens of thousands of downloads: on Douyin and Bilibili, a batch of "one GitHub gem a day" software resource accounts kept reposting it as a "frame-rate booster," one after another. Below: what it is, how this traffic came about, who it's competing with, how you could get your own open-source project starred the same way, and what it might yield.

A 56-second screen recording

At 9 a.m. Beijing time on Nov 2, 2026, the Douyin account 「吃一颗豆!」 posted a 56-second video. The account has only 199 followers. The caption reads:

The mountain that boosts game frame rate the most: Pavise, one-click for beginners, supports competitive mode, custom mode, GPU policy tuning and system status detection. Once enabled, it can temporarily schedule CPU, memory, disk and background processes, and restores settings automatically on exit. Very useful.

The hashtags were #豆包工作 #豆包 #游戏帧率 #显卡 #三角洲帧率. By Nov 6, the video had 20.6k likes, 17k saves, 7,089 shares and 489 comments (Douyin video page, 11-06). The save count is close to the like count: viewers wanted to come back and download it.

The comments are barely praising it. Mostly people asking: "Will this get my Ducking account banned?" "Laptop player on Where Winds Meet, can I use this? Heard it makes low FPS more stable 4050" "Can I use it for PUBG? I've got a 4060 laptop" "What's the difference from Process?" "Is it paid?" The top-liked comment (46 likes) is a warning: "Seems to clash with Little Yellow Duck. After using it, opening Little Yellow Duck again makes things lag." Others say "Used it, still dropping frames."

Over the next few days, the same line "the mountain that boosts game frame rate the most" was copied word for word: from 11-03 to 11-05, a dozen-plus Bilibili accounts such as "清澈工具箱", "光隐资源库", "极刻推荐", and on Douyin "豆爆丁" (6,130 likes), "叫我姜姜i", "大鱼丸子", etc., posted similar 40 to 60-second screen recordings (Bilibili and Douyin search for "pavise", 11-06).

On GitHub you can see the traffic landing: 11-02 shipped v3.2.0.2, and by 11-06 the zip and exe downloads together came to about 17k, while the checksum file the in-app update check pulls was fetched 6,370 times. 11-06's freshly shipped v3.3.0.7 got 2,406 downloads on day one (GitHub Releases, 11-06). Meanwhile stars went from 11-01's 603 to 11-06's 828: +241 over the last 7 days, versus just +24,11-05 in the first 7 days, and +78 on a single day (GitHub daily star counts). Downloads run dozens of times ahead of stars: the people coming in are players, most don't have a GitHub account, and don't know what a star is.

What it is

A C# Windows utility under GPL-3.0. During gaming, it suppresses background processes, gives the game exclusive CPU cores (distinguishing AMD's CCDs and Intel's P/E cores), moves device interrupts off the game's cores, and sets GPU driver, memory and power policies per game; when the match ends, all changes are reverted. The README says it does not inject and does not collect data; for anti-cheat systems like ACE, EasyAntiCheat and BattlEye it only handles user-mode services and "does not touch kernel anti-cheat drivers."

Beginners get two one-click presets ("author's own setup" and "extreme"); tinkerers get per-game configuration, frame-time and 1% Low display, and logs. There's also the official site pavise.club and tutorials, and the author ships a release roughly every two weeks (10-27, 11-02, 11-06). The README's last paragraph reads: "Pavise is permanently free, donations are entirely voluntary. If you paid for Pavise through another channel, please request a refund from the seller or trading platform." People are already reselling it.

The author, bdth, has 108 followers on GitHub and has built Chinese-community mini programs, a ChatGPT wrapper, and a ComfyUI management platform (788 stars), with experience taking Chinese projects to a few hundred stars.

Where the traffic came from: not GitHub circles, but software resource accounts

Developers barely talked about it: Hacker News, Reddit, V2EX and linux.do returned no discussion, and on X only one resource account, August, posted about it; with the repo under 60 days old, it couldn't make GitHub's trending list. Distribution came in two waves, both on short-video platforms:

Why would resource accounts post it? These videos commonly carry hashtags like #豆包 #豆包工作 or #百度搭子 #办公就用dumate, which are creator activity tags for AI products; resource accounts monetize through event traffic, rewards and follower growth (reward rules unverified). A free, open-source, uncomplainable piece of software is a zero-cost, zero-risk topic for them. Pavise itself spent no money; it was carried along by someone else's monetization chain; no evidence was found that the author ran paid promotion.

Why it: the tech isn't novel, the packaging is right

CPU pinning, interrupt affinity and power plans have long been done by veteran gamers:

English equivalents do similar things but stall at a few hundred stars, gaining a few a week. What Pavise adds is packaging that lets resellers push it and beginners click it: Chinese UI, one-click presets, "only active during matches, auto-reverts when the match ends", "no injection, doesn't touch kernel anti-cheat", tailored to Delta Force, Valorant, LoL, WeGame and other domestic games, plus a ready-made recommendation line: "free, open-source, replaces the paid one". 40 seconds of screen recording covers the whole pitch, and the "will it get me banned" worry that fills the comments is answered right in the README.

If it were you

In one line: the thing to learn is this distribution chain: "turn an existing hardcore English-language tool into a one-click Chinese version, and let software resellers on Douyin and Bilibili push it for you." But know what this chain mainly brings: downloads, not stars. Tens of thousands of downloads yielded only a couple hundred stars. Another game optimizer: Pavise already holds that spot.

Who it suits: developers who can build a decent UI for a Windows or desktop utility, willing to write "safe, free, reversible" into the README and the interface, and fine with users mostly not being developers.

Who it doesn't suit: anyone aiming for thousands or tens of thousands of stars (this chain delivers hundreds); anyone who only wants reputation within GitHub, X or HN circles (nobody there knows it); anyone who can't stand comment sections full of "will it get me banned" and "does it even work" and having to answer beginners' questions endlessly.

How to do it, from most to least reliable:

1. Switch category: look for hardcore tools that exist in English but have no one-click Chinese version: network latency optimization, GPU driver cleanup, storage cleanup, screen recording, keyboard/mouse macros, monitor color calibration, and similar, where the mainstream has a pain point and the effect can be shown in 40 seconds. First check whether this category has an English library like CPUSetSetter that does the right thing but nobody pushes.
2. Do what Pavise didn't: Pavise's gap is "visible effect." The biggest questions in the comments are whether it works, whether it conflicts with Xiao Huangya, and whether it gets you banned. A tool that automatically compares 1% Low before and after a session and publishes a per-game compatibility list, or a version aimed at laptops (the comments are full of 4050 and 4060 laptop owners), is an angle that can catch the same audience.
3. Change language: build the same thing as an English or other-language version and pitch it to "free PC tool" accounts on TikTok and YouTube Shorts. Whether the Chinese reseller mechanism has the same scale overseas is unverified; you'd be betting on it yourself.

How to get users: record your own 40–60 second screen capture with a quotable line (Pavise got copied for "the mountain of highest frame-rate boosting"), post it on Douyin and Bilibili, and tag current popular creator campaigns; organize the tutorial, screenshots and download link, and message "one GitHub gem a day"-style reseller accounts so they can post it right away. Host downloads on GitHub Releases and put a line at the top of the README asking users to star the repo, or only about one in a hundred of tens of thousands of downloads will come back. You need a website and in-app updates too, so that when traffic arrives there's a new version ready.

When to move before it's gone: watch whether the activity tags resellers are chasing still exist. When campaigns like the Doubao or Baidu Dazi contributor programs stop, resellers' incentive to pick topics weakens. Meanwhile game optimization has already been run through by Pavise, so if you switch categories, do it before resellers exhaust the "GitHub gem" series.

Signs you misjudged: ① over the next few weeks Pavise's downloads fall off while stars stay at eight or nine hundred, and you see a wave of bans and frame-drop complaints, meaning this kind of "boost" tool doesn't keep users: resellers can push downloads but not reputation; ② another Chinese one-click tool pushed by resellers gets just as many views yet barely gains stars, meaning stars were an accidental spillover and this chain isn't worth it for stars; ③ conversely, someone uses the same playbook in another category and gets several hundred stars in a week, meaning switching categories works and you should move fast.

---

Data and sources: the GitHub repo dulaiduwang003/Pavise-Game's README and Releases download counts (11-06); GitHub daily star counts (10-16 to 11-06); the copy, engagement numbers and comments on Douyin video 7703372834949162459 (11-06); Douyin and Bilibili searches for "pavise" (11-06); the X "Weekly Resource Roundup" 8-09 post; download sites tenlonstudio, 61ku, cymao, ckbf88, and the CSDN and Sohu reprint pages; competitors' GitHub pages and Bilibili videos (11-06); the official site pavise.club. The relationship between resellers and creator campaigns, and the campaign reward rules, are unverified; there is no reliable data on actual optimization results or ban risk.