An open-source tool that gives coding agents like Claude Code and Codex a language and runtime for making video: drop in a viral video and the agent breaks it into an editable workflow of footage, word-aligned subtitles, B-roll, and effects, then generates hundreds of variants with one command. The hosted model service is called HypiHub and is charged by subscription and credits.
In 2026-09 it open-sourced on GitHub an agent tool that lets Claude Code/Codex clone viral videos. Concentrated coverage from the top spot on Trendshift, AI media, and Chinese WeChat official accounts brought 20,000+ stars, and it monetizes through subscriptions and credits for the hosted model HypiHub, reaching $27K in revenue in the very first month.One-line growth story, compiled by us from public sources; speculation is marked separately below.
The main engine is open-source community spread: after open-sourcing on GitHub it hit the top of Trendshift, and overseas AI media and Chinese WeChat accounts followed with reviews; Similarweb has no data on channel shares.
Ranking first on the Trendshift overall chart two days in a row was the turbo boost; the lubricant is subscriptions and credits, and TrustMRR records 782 subscriptions.
By estimate, about 1.8 payments per 100 visits (including renewals); the visit count is estimated from global rank.
Model: Lenny's racecar growth framework. Engine = open-source community spread (GitHub trending charts, secondary spread by media and bloggers, communities); turbo boost = No. 1 on the Trendshift overall chart two days in a row; lubricant = subscriptions and credits.
Real data: 1,126 payments, $26,844 in revenue and 782 subscriptions come from TrustMRR (the payment month is last month); the Similarweb public page has only the top search keywords (15 branded terms including hypit, hypit ai and hypihub) and social platform names (X, LinkedIn), with no channel shares, countries or bounce rate. The 62,100 total visits is estimated from global rank and is not a real figure.
Supporting evidence: The Trendshift page records the repository at No. 1 on the overall chart on two single days and about 26,700 stars; the repository had about 1.3K stars when AlphaSignal published; a Menon Lab blog post and a Jiamu Weilai Pai WeChat article reviewed it one after another; the README lists partnerships with the Discord, Telegram and LINUX DO communities; the founder self-reported the channels on TrustMRR as Twitter and Discord.
Inference: Total visits are an estimate, so payments as a share of visits (about 1.8 per 100) is only a reference, and the channel circle carries ≈. The split: charts 24,000 + media 18,000 + communities 12,100 + subscribers returning 8,000 = 62,100. The 5% for charts, 10% for media, 20% for communities and 50% continuing to browse, and this split, are all guesses. Most open-source users go straight to GitHub and never reach the official site, so official site visits are only a small part of them.
Real payments synced by TrustMRR, in USD, by calendar month, renewals included.
“"Star on GitHub"; "Fork, modify, self-host". Homepage: "$0 open-source core". Pricing page lists no free plan. Discord link present.”Quote from the site
New-user source: Other (Direct / referral / email / paid). There is no Similarweb channel breakdown (the site is too new, with only 15 keywords, all brand terms like hypit, hypit ai, hypihub, and hypit github; social is only X and LinkedIn). New users come mainly from the GitHub open-source release climbing to No. 1 on Trendshift (No. 1 overall on 2026-09-15 and 09-16), overseas AI media (AlphaSignal, Menon Lab), and reviews on Chinese WeChat official accounts and news sites (甲木未来派, AI 小集, and others); the README also lists a partnership with the LINUX DO community, and there are Discord and Telegram groups. The founder self-reported Twitter and Discord as channels on TrustMRR. Existing users stay through credit consumption and subscription renewals.
| Attributable new users, last month | Share |
|---|---|
| Other | 100% |
Direct + brand search 0%. Direct and brand search count as viral only when the site shows sharing mechanics; the largest share wins, under 5% means Other.
English
X, LinkedIn
15 keywords · brand ~0% · organic 100% / paid 0%
| Similarweb top 5 keywords |
|---|
| hypit |
| hypit ai |
| hypihub |
| hypit clone video |
| hypit github |
Non-brand queries: hypihub
About $26,844 last month (real payments synced by TrustMRR), #71 of 563 TrustMRR sites; the month before was $0.
New-user source: Other (Direct / referral / email / paid). Mostly referrals, email or paid, or mostly direct with no clear source There is no Similarweb channel breakdown (the site is too new, with only 15 keywords, all brand terms like hypit, hypit ai, hypihub, and hypit github; social is only X and LinkedIn). New users come mainly from the GitHub open-source release climbing to No. 1 on Trendshift (No. 1 overall on 2026-09-15 and 09-16), overseas AI media (AlphaSignal, Menon Lab), and reviews on Chinese WeChat official accounts and news sites (甲木未来派, AI 小集, and others); the README also lists a partnership with the LINUX DO community, and there are Discord and Telegram groups. The founder self-reported Twitter and Discord as channels on TrustMRR. Existing users stay through credit consumption and subscription renewals.
Open-source core $0, self-hostable, no watermark, no seat fees; hosted service subscriptions $9.90 / $39.90 / $99.90 per month ($118.80 / $478.80 / $1,198.80 billed annually); credit packs $15/2,000, $35/5,000, $65/10,000, $130/20,000. The hosted service pricing page has no free tier
Dossier confidence: medium. Parts 1 and 2 come from reading the site and Similarweb data; part 3 from public reporting, founder posts, Wayback snapshots, WHOIS and similar, with speculation marked.
The numbers: revenue comes from TrustMRR: real payments synced automatically from the founder's Stripe or other payment account, summed by calendar month, renewals included, in USD. Site visits come from the Similarweb public page; where it shows none, they are estimated from global rank or the TrustMRR payment count, as noted on the page.
Data through last month. Page updated 2026-10-31.