Similar tools: how they work and what we took
Research from November 2026, based on public READMEs, documentation and product pages. Nothing behind a login was checked. Prices and features change, so check the linked pages before relying on them.
The common pattern
Almost every tool in this space does the same three things:
- Find outliers. It compares a video's views with what that creator usually gets. The baseline is the creator's median (the sturdier choice) or their mean, and sometimes it's matched by age.
- Pull the hook and transcript. The hook is the first ~3 seconds, and it's the part people copy.
- Turn hits into work. An LLM breaks down why the video worked and writes adapted scripts, and the hits go into saved collections ("swipe files") and alerts.
Open source
viral-shorts MCP (Xeron2000/shorts)
- Data: YouTube Data API v3 only.
search.listcosts 100 quota units;videos.listandchannels.listcost 1 each. - Scoring:
- VPH = views ÷ hours since publish. 1K VPH means potential, 5K high, 10K "super viral".
- Engagement = (likes + comments) ÷ views.
- Viral score = VPH × a time weight × (1 + an engagement boost), with newer videos weighted higher.
- Output: a Markdown table from an MCP client, on demand. No scheduling.
viral_short_research_tool (wisnuhid7)
- Data: YouTube Data API v3; it deliberately avoids scrapers.
- Scoring: velocity (views/hour) counts 45%, reach (views ÷ subscribers) 30%, engagement 25%. Videos under 3 hours old get a confidence penalty.
- Extracts: GPT-4o-mini splits each video into hook (the first 3 s), body and CTA, plus presentation style. The result is cached per video and per prompt version.
- Output: suggested keywords taken from the tags and title words of top videos; saved collections with notes; CSV and Markdown export; a quota meter.
ViralMint (openclaw-easy/ViralMint, AGPL-3.0)
- Data: the YouTube API; TikTok and Douyin through TikHub, a paid API (its cookie fallback breaks the platforms' terms); yt-dlp; Google Trends.
- Scoring: outliers at 3–20× the channel's baseline, plus an "AI virality" score. The formulas aren't published.
- Extracts: local faster-whisper transcripts. An LLM pulls out the hook, structure, tone, retention risks and suggested titles.
- Output: alerts via Telegram, Slack or Discord. It also cuts and renders clips, which makes it a full production pipeline.
Commercial
| Tool | Platforms | How "viral" is measured | Beyond numbers | Price |
|---|---|---|---|---|
| 1of10 | YouTube (+ Shorts) | views ÷ the channel's typical views; shows ~6–100× | Chrome overlay, tracked channels with notifications, bookmarks and folders, AI title/thumbnail/idea generators | $29–69/mo |
| vidIQ Outliers | YouTube | views vs the channel's average over a similar timeframe (age-matched); bands <2×, 2–5×, 5–10×, >10× | filters for score, views, subscribers, VPH, length, publish window | subscription |
| TubeLab API | YouTube | returns both zScore and averageViewsRatio, with positive and negative outlier flags |
semantic search, type=short, LLM labels (faceless, quality), a transcript endpoint |
credits |
| Virlo | TikTok, YouTube, Instagram | outlier = views ÷ the creator's median of up to 100 recent videos; 1.5× above average, 3× viral, 10× mega viral; weighted = ln(outlier) × ln(median); also a percentile and z-score | trending hashtags, sounds and hooks; scheduled niche "research agents"; webhooks; MCP | $0.50 per check |
| Sandcastles.ai | IG, TikTok, YT Shorts | "top 1%" outliers (formula not disclosed); tracks competitor channels every 6 h (1 h on top tiers) | transcripts, breakdowns of hook, story, format and style, hook templates, script writer, collections, MCP | $49–499/mo |
| Apify TikTok hook scraper | TikTok | views ÷ followers, outlier at 3× or more; engagement = (likes + comments + shares + saves) ÷ views | parses TikTok's own captions: anything in the first 3.5 s becomes openingHookText, then body and CTA; hook archetype; words per minute; sound ID |
~$0.01–0.025 per post |
| TikTok Creative Center | TikTok | trending hashtags, songs, creators and videos by region, industry and period | trendlines and audience data; no public API or export | free (login recommended) |
| Exolyt, Pentos | TikTok | hashtag and sound trends, competitor dashboards (data source not disclosed) | none noted | subscription |
Other sources we checked:
- Google Trends: the official API is an alpha that you have to apply for (announcement). The community
pytrendslibrary is reported broken. - Instagram Graph API hashtag search: official, but it needs the Public Content Access feature and is limited to 30 unique hashtags per 7 days (docs).
omivibe.com (checked because it was suggested)
omivibe.com is a B2B lead-generation and cold-email product, not a video or trend tool:
- It finds leads from Google Maps and a B2B database.
- AI writes a personal email to each lead, sent from your own Gmail or Outlook.
- It writes and posts LinkedIn content automatically.
- It has an MCP server whose tools run as async jobs and check quota before acting.
- Price: $49–349/mo.
The useful ideas it gave us: drafts that come straight out of findings, a confirmation step before anything is published, budget warnings at 80% of a cap, and an MCP server with a start-a-job-then-poll pattern.
What viral-scout took
| Idea | From | In viral-scout |
|---|---|---|
| Outlier = views ÷ the creator's median | Virlo, 1of10, vidIQ | × usual column |
| Bands of 1.5× / 3× / 10× | Virlo | ↑ / ⚡ / 🔥 labels |
| Rank by ln(outlier) × ln(baseline) | Virlo | Outliers section order |
| Penalty for very new videos | wisnuhid7 | ⏳ under 6 h; per-day floored at 6 h |
| Hook = the first ~3–4 s of captions | Apify hook scraper, wisnuhid7 | Hooks section, the hook field |
| LLM breakdown + adapted scripts, cached | wisnuhid7, ViralMint, Sandcastles | the /viral-scout skill (Claude Code, no extra key); transcripts cached in SQLite |
| Swipe file / collections | 1of10, Sandcastles | save → Obsidian note + swipe.base |
| Tracked channels + notifications | 1of10, Sandcastles, ViralMint | watch list + Telegram alerts |
| Keyword suggestions from hits | wisnuhid7 | suggested hashtags |
What we do that most of them don't:
- Growth between runs ("rise/day"). Most tools only see a snapshot.
- Per-business profiles, including the business description the AI uses.
- Free on YouTube, and output into Obsidian.
Added in round 2 (after this research)
| Feature | From | In viral-scout |
|---|---|---|
| Hook / body / CTA split of the transcript, words per minute | Apify hook scraper | cta (last 5 s), wpm; shown in the report, the dashboard and the JSON |
| Percentile among the creator's recent posts | Virlo, TubeLab | creator_pct ("beats 87% of their posts") |
| Local Whisper transcripts for Instagram/TikTok audio | ViralMint | transcribe.whisper (faster-whisper, CPU, int8; audio decoded by the system ffmpeg) |
| On-screen text of silent hooks | Sandcastles, TubeLab (vision LLM) | thumbnails cached in data/frames/; the skill reads them with Claude's vision, so no OCR install is needed |
| Quota meter and budget warnings | wisnuhid7 (quota meter), omivibe (80% alerts) | usage per run (YouTube API units, Apify results, page requests, Whisper seconds), warnings at 80% of budget: caps |
| MCP server with start-then-poll jobs | Virlo, Sandcastles, omivibe | python -m viral_scout.mcp_server (11 tools) |
| Web dashboard with collections and status board | 1of10, Sandcastles | web/ (Next.js), local with actions, or read-only on Vercel |
Refresh: newer competitors (November 2026)
New tools found:
- Mysocial Outliers: creator median plus a topic median, top comments, an MCP server.
- Trendhaus: TikTok "Post DNA", a hook library, comment mining, trend stage.
- OutlierKit: 3–5×, 5–10× and 10×+ bands, AI review of the first 15 s.
- Satura: an emerging-channel radar and a WebSub channel tracker.
- Viewstats, ShortsMonkey MCP.
- youtube-outlier-finder: finds small channels with Shorts at ≥10× their subscribers.
- ig-radar-plugin: a Claude Code plugin with an own-post postmortem.
Full list with links: references.md.
Added in round 3
| Feature | From | In viral-scout |
|---|---|---|
| Comment mining: questions and top comments under the hits | Trendhaus, Mysocial | comments (YouTube API commentThreads, or yt-dlp public comments); report section "What viewers ask" |
| Topic-relative score (× the niche median) | Mysocial | vs_topic, from data we already have |
| Own-post postmortem: your posts against your own usual | ig-radar-plugin | profile own: {youtube: [...], urls: [...]}, a "Your posts" tab |
| Emerging creators (small channel with a clear outlier) | Satura, youtube-outlier-finder | 🌱 emerging flag (≤50K followers and ≥3×) |
| Age-matched baseline (views/day vs the creator's views/day) | vidIQ | vs_pace, with the YouTube API key (2–3 units per creator) |
| Trending hashtags | TikTok Creative Center, Exolyt, Virlo | Apify trends actor (APIFY_TRENDS_ACTOR) and Instagram hashtag top_media (Graph API, 30 a week) |
| Local competitors | (none; asked for by the user) | competitors command: Google Places API (New), plus social links from each business's homepage |
| Search across everything collected | TubeLab (semantic), Sandcastles (library) | search (BM25 over titles, hooks, CTAs, transcripts), the Search page, the MCP search tool |
| Setup page for keys, with live tests | the paid tools' settings pages | dashboard /setup (local only; stored in .env; masked) |
| New research from a description | Virlo research agents | dashboard /new, new command, MCP create_project |
Still not built (ranked)
- Semantic (embedding) search. BM25 covers keyword search. Embeddings would need a local model.
- Scene-by-scene breakdown (several frames per video). We read one thumbnail. Downloading whole videos is limited by platform terms, so this should stay with your own or licensed posts.
- YouTube WebSub push for watched channels. Free, but needs a public callback URL, for example a Vercel route.
- Instagram own-post insights through the Graph API Media Insights (own business account only).
- X / Threads coverage. Low value for short-form video.
Earlier backlog (now done)
- Trending sounds and hashtags. TikTok Creative Center has no export, so this needs an Apify trends actor (paid). For Instagram, Graph API hashtag search (30 a week).
- Age-matched baseline (vidIQ): compare views per day with the channel's median views per day at a similar age. It needs per-video dates in channel listings, which costs more requests.
- Local competitors: use the official Google Places API to find nearby businesses in the same niche, then look up their public channels as watch-list candidates.
- Semantic search over saved transcripts (TubeLab): an embeddings index of the swipe file.