The one-glance read on who they are and how they grow. Each point is verifiable from the receipts above.
A browser-based AI video and image platform bundling 100+ tools (face swap, lip sync, image-to-video) for creators and marketing teams, out of Y Combinator's Winter 2024 batch.
Organic-led and launch-led: free tool pages, YouTube tutorials, and a founder building in public, with Google search ads as a defensive layer underneath.
~1.7M estimated monthly visits (+28% over the last 3 months); a 169,364-follower owned Instagram; an estimated ~400-ad Google library with at least 44 ads still active as of 2026-07-13; revenue is not company-disclosed, third-party trackers estimated it at roughly $1.7M ARR in 2025.
Founder Runbo Li recounted that "an investor who passed on us emailed saying he saw our video at a Mavs game," after which he traced it to the Dallas Mavericks' own in-game hype reel, built with Magic Hour.
An ad reading "AI Video & Image Generation API, Try Magic Hour today" signals a developer/API-positioning test running alongside the consumer free-tool ads, active as of 2026-07-13.
Magic Hour's own AI outputs, sports clips, music videos, tutorials, are the marketing content, so content production doubles as product demonstration with no separate creative budget required.
The order the channels came online. Sequence is strategy: what they did first, and what they layered on once demand existed.
Estimated demand, the channel split behind it, and the keywords and referrers doing the work. Directional modeling, not audited analytics.
| Keyword | Volume | Weight | CPC |
|---|---|---|---|
| face swap | 96k | $0.40 | |
| face swap ai | 66k | $0.42 | |
| ai face swap | 56k | $0.47 | |
| magic hour ai | 13k | $0.57 | |
| magic hour | 21k | $0.57 |
Magic Hour draws an estimated ~1.7M monthly visits, up 28% over the last 3 months. Search Organic leads at 44.7%, Direct follows close behind at 41.4%, together accounting for the large majority of traffic, and Referrals add another 5.9%; the rest, social organic, generative-AI referrals, email, and paid search and social, totals about 8%. That Direct share sitting almost as high as Organic Search points to a sticky, bookmark-and-return user base rather than one that rediscovers the tool through search each time.
The keyword base is dominated by high-volume, high-intent transactional terms: "face swap" (95,800 searches/mo) and "face swap ai" (65,540 searches/mo) dwarf the branded term "magic hour" (20,740 searches/mo), meaning most of the demand they're chasing is generic tool-category search, not brand recall. Top referrers include ycombinator.com (the YC alumni network surfacing them to other founders and investors), and two AI-tool directories, otechworld.com and toolify.ai, which suggests a chunk of inbound traffic comes from directory placement rather than editorial press. Among stated competitors, heygen.com is the best-funded AI-video rival, remaker.ai is a direct face-swap competitor that Magic Hour targets by name in its own content (see below), and aifaceswap.io is a narrower niche player in the same category.
The specific pages earning their organic search traffic, and the pattern behind why they rank. Adapt the format, not the topic.
Magic Hour's organic traffic concentrates almost entirely on individual tool pages, not blog content: the video face-swap page pulls ~84,942 organic visits/mo (32.8%) and the image face-swap page ~67,845 (26.2%), both ranking #1 for their exact-match terms. The pattern is a working free tool sitting directly behind the ranking keyword, so the search result and the product experience are the same page. The one blog-format exception is a comparison post targeting a competitor's brand name, "remaker ai face swap," which pulls ~52,416 visits/mo (20.3%), ranking #6 for "remaker ai" itself, a page built specifically to intercept a rival's branded search. The homepage and a third tool page (AI clothes changer) round out the top five at single-digit shares.
Not traffic share. How much weight the growth system actually puts on each channel, with a one-line read on the role it plays.
For founder-led SaaS the breakdown shifts from ads to traction: where the first users came from, how the founder grows it in the open, and the compounding organic surface.
Magic Hour's first users came from AI-generated content the founders posted before the product existed, followed by two Show HN attempts in the same summer that only broke through modestly.
In July 2023 the team posted a batch of AI-generated music videos (Tyler the Creator, Kanye West, The Beatles) and sports clips (an AI-generated Damian Lillard shot) to YouTube under the Magic Hour name, seeding an audience before there was a public signup flow.
A first Show HN post on 2023-07-30 ("Create your own AI videos with Stable Diffusion") drew 3 points and no discussion; a second post two days later, "Show HN: Create AI Videos Easily," stated plainly "That's why I built Magic Hour," and drew 7 points and 2 comments, the actual naming of the product to the HN audience.
The clearest inflection in the evidence is not a launch spike but the unorchestrated Mavericks placement noted above, the kind of organic distribution the company did not plan for.
A third Show HN post surfaced on 2026-03-02 ("Audio-to-Video with LTX-2," 23 points), a return to the same launch platform roughly two and a half years later tied to a new model integration rather than a new product, showing the founder treats Show HN as a repeatable checkpoint rather than a one-time event.
Runbo Li's personal X account is the entire visible build-in-public motion, small in reach but consistent in cadence and specific in what it shares.
At 312 followers and 842 posts, Li's reach is modest, but he posts recurring, specific build detail rather than generic updates, including a note that a feature "took 3 months and 10,000 throwaway videos" to ship.
His best-performing post (82 likes, 11,076 views) is the Mavericks pregame-video story above, and a separate post notes NBA star Kevin Durant sharing a Magic Hour-made video, both framed as unsolicited third-party validation rather than self-reported growth numbers.
A July 2025 post about the team taking 1st, 2nd, and a GPU prize at a ComfyUI (an open-source AI image/video workflow tool) hackathon doubles as recruiting and credibility content for a technical audience.
The evidence shows no personal LinkedIn or Reddit presence for the founder, so the build-in-public motion is concentrated on X alone rather than spread across community platforms.
Search Organic is the largest single slice of Magic Hour's traffic mix at 44.7%, and the strategic layer behind it is a one-page-per-feature build: dozens of standalone tool pages (face swap, clothes changer, image-to-video, lip sync) each targeting its own exact-match keyword rather than one general homepage.
Each of the 100+ tools gets its own indexable, working landing page, so the product surface and the SEO surface are the same asset, compounding as new tools ship.
The remaker-ai comparison page covered above is a deliberate play to intercept a named rival's branded search rather than only generic category terms.
All identified paid activity runs through Google Search (no Meta or TikTok ad evidence), split between "proven winner" bold-claim text ads on free-tool terms ("100% Free AI Image Generator," "AI Video Face Swap, Free, Instant") and one before/after face-swap creative built around a Robert Downey Jr./Tom Cruise Iron Man comparison; this reads as defensive brand- and free-tool-term capture layered on top of the organic engine, not a cold-audience acquisition channel.
This founder-led SaaS also runs paid acquisition. Here are the live ads doing the work, each with the X-ray and a play you can adapt.

Why it works. Doubles down on zero cost in both headline halves so price-sensitive searchers evaluating multiple AI art tools see no reason to hesitate.
Headline with a bold claim: '100% Free AI Image Generator - Create Free AI Art' Repeat your strongest value word (free, instant, unlimited) in both headline halves instead of stating it only once.

Why it works. Targets a specific technical use case (lip sync) rather than generic AI image claims, catching searchers who already know what they need.
Headline states: "Create high-quality, realistic Lip Sync animations for free" Run a separate search ad for each specific feature you offer (like lip sync) instead of one generic AI tool ad.

Why it works. Repeats the exact long-tail search phrase (clothes changer) twice, maximizing quality score relevance for that niche keyword.
Headline offers a free tool: 'AI Clothes Changer: Free - Try Our AI Clothes Changer' For a niche feature, repeat the exact phrase people search for verbatim in both headline slots instead of varying the wording.

Why it works. Bids on a competitor's brand name (OpenArt) so its own ad appears when people search for a rival tool, redirecting that intent to magichour.ai.
Visual demonstration of product output (image of astronaut) combined with a clear product name (OpenArt - AI Image Generator). Bid on your closest competitor's exact brand name as a search keyword so your ad shows up next to or above their own listing.

Why it works. Pairs the brand name with two professional headshot outputs so viewers instantly grasp the use case without reading copy.
Displaying before-and-after style images (implied by the small circular images at the bottom of the main headshots) with a LinkedIn logo. Show two polished example outputs side by side for your most popular specific use case (like professional headshots) with no other copy needed.

Why it works. Uses two instantly recognizable actors in the same iconic role to make an abstract feature (face swap) concrete in half a second.
Visual demonstration of product capability: The main image shows a 'before' (Robert Downey Jr. as Iron Man) and 'after' (Tom Cruise as Iron Man) with a small source face (Tom Cruise) and an arrow, illustrating the face swap feature. This is accompanied by the text "AI Face Swap - Free & Easy". Demo a swap or transformation feature on two widely recognizable faces or references so the effect reads instantly without explanation.
The channels are not separate. They are one system where each stage feeds the next. Here is the read, then the plays to run tomorrow.
Magic Hour's loop starts with a free tool page ranking for a specific transactional keyword, converts on the spot since the demo and the product are the same page, and compounds through owned content (YouTube tutorials, the large Instagram account) rather than paid spend.
A visitor searching "face swap" or "face swap ai" lands directly on a working free tool, collapsing the usual gap between ad click and product trial into a single page.
YouTube tutorials and the 169K-follower Instagram account continuously restock top-of-funnel demand without incremental ad cost, since the content is the product's own output.
The founder's personal audience (312 X followers) is far smaller than the brand's owned channels, so almost all the compounding equity sits in owned product pages and the Instagram account rather than in founder reach, a fragile spot if either channel's algorithmic favor shifts.
free-to-paid conversion rate, churn, CAC, and confirmed current funding status (public trackers disagree on whether Magic Hour has raised outside capital at all).
The proofMagic Hour posted AI-generated music videos and sports clips to YouTube in July 2023, months before running its Show HN launches, seeding an audience before there was a signup flow.
The adaptationIf your product generates any visible output, video, image, document, dataset, publish raw examples of that output to the platform your target user already watches before you open signups. First step: pick 5-10 outputs that look good even without context, post them with a one-line caption naming the tool, and track which ones get unsolicited comments asking "how did you make this," that's your validated hook for the real launch post.
Cost: $0 · Time to signal: days · Works pre-PMF: yes, provided the raw output is genuinely watchable without explanation.
The proofMagic Hour posted three separate Show HN launches years apart (2023-07-30, 2023-08-01, and 2026-03-02), each tied to a real new capability rather than a repost of the same pitch.
The adaptationTreat your community launch platform as a recurring checkpoint tied to genuine capability jumps, not a one-time event. First step: the next time you ship a change a technical user would care about, write a short honest build post about what changed and why it was hard, and post it to the same community you launched on originally.
Cost: $0 · Time to signal: days · Works pre-PMF: yes, conditional: only relaunch on genuine capability jumps, repeated self-promotion without new substance reads as spam and can get flagged.
The proofAn AI-generated Damian Lillard clip and a Mavericks in-game hype video, both made with Magic Hour, reached audiences the company never directly targeted, including a Mavericks staffer who put one on the arena screen.
The adaptationGenerate content in your tool using source material a specific niche audience already shares (a sport, a fandom, a local scene) and hand it directly to the accounts or teams that community follows, not just post it and wait. First step: pick one niche your product's output fits, make three pieces of shareable content in that niche's native format, and send them directly to two or three community or team accounts likely to repost.
Cost: $0-under $500 · Time to signal: weeks-months · Works pre-PMF: conditional, only if your tool's output is visually or audibly shareable on its own, and be cautious using real people's names or likenesses without clearance, that carries publicity-rights and platform-policy risk at any scale.
The proofMagic Hour's face-swap tool pages, each targeting its own exact-match keyword, drive over half of its organic traffic and rank #1 for their terms (covered in Organic Content above).
The adaptationIf your product bundles multiple distinct capabilities, give each one its own indexable page with a working demo embedded, not a single feature list behind one signup wall. First step: pick your single most-searched capability, check its monthly search volume, and build one standalone page with a functioning free version of just that feature.
Cost: $0-under $500 · Time to signal: months · Works pre-PMF: conditional, you need at least one feature stable enough to demo without a login.
The proofThe remaker-ai-face-swap comparison page covered above is Magic Hour's third-largest organic traffic source, built specifically to intercept a rival's branded search.
The adaptationIdentify the one competitor your prospects search for by name, and publish a direct, fair comparison page targeting "[competitor] alternative" or "[competitor] vs" queries. First step: check search volume for your top competitor's name plus "alternative," write an honest comparison that acknowledges what they do well, and publish it on your own blog.
Cost: $0 · Time to signal: months · Works pre-PMF: yes, provided a competitor with real branded search volume exists; be aware some jurisdictions and ad platforms restrict using a competitor's trademark in paid creative, so keep this to organic content. Not transferable at an earlier stage: Magic Hour's 169K-follower Instagram account and its unsolicited placement in an NBA arena's in-game rotation are compounding assets built on top of an initial viral moment and roughly two years of continuous output, not something a pre-launch founder can replicate directly; adapt the underlying mechanics above, not the current scale.
Every morning we take one company that is actually growing and break down where its customers come from: the ads still running after a year, the channel doing the real work, and the play you can run this week.
Systemaic · directional intelligence. Traffic, spend, and reach figures are SimilarWeb-style estimates and qualitative reads of public data, not audited numbers. Built on real public receipts.
Live ad libraries: google ad library
Launch archives: Hacker News: Show HN: Audio-to-Video with LTX-2 · Hacker News: The Magic Hours: The Films and Hidden Life of Terrence Malic · Hacker News: Show HN: Create AI Videos Easily · Hacker News: 24 magic hours to SearchSpin - Story of a Yahoo Hack Day Ind
Traffic, spend, and revenue figures are estimates as noted in the report; the links above are the primary public artifacts.