Trend detection across social, forums and video
A normalised stream of public posts, comments, reviews and video transcripts across 40+ platforms. Topic models, sentiment, virality scores and entity extraction included.
Processing 1.8B public posts per month with sub-90-second ingestion on priority sources.
{
"query": "brand:whoop",
"window": "6h",
"virality_score": 74,
"trend": "rising",
"top_platforms": ["tiktok", "reddit", "youtube"],
"posts_sample": [
{
"platform": "tiktok",
"author_handle": "@fitsamuk",
"text": "6 months on the Whoop 4.0 - honest sleep data review...",
"entities": [{"type": "brand", "value": "Whoop"}],
"sentiment": 0.62,
"reach_estimate": 184000,
"posted_at": "2026-04-24T06:12:00Z"
}
]
} Built for the jobs data teams actually do.
Beyond the usual platforms
Reddit, TikTok, YouTube, X, Threads, Instagram, Bluesky, Discord (public), Twitch chat, plus 30 niche forums and review sites.
Trend scores, not volume graphs
A virality score per topic blends velocity, reach and cross-platform spread. Catch inflections, not just yesterday's peaks.
Video included
TikTok and YouTube transcripts are ASR-processed and indexed, so text search works across spoken content too.
Entities and sentiment
Every post tagged with brands, products, people and places, plus calibrated sentiment — not just positive/negative/neutral.
Integrate in an afternoon.
Describe what to watch
A brand, product, keyword set, competitor handle or seed URLs. We build a live query and show sample matches within minutes.
Stream or query
Pull matched posts and trend scores via REST, or subscribe to a websocket/webhook for real-time alerting.
Act on inflections
Trend-score crosses and sentiment shifts trigger alerts into Slack, email or your CRM — without you watching a dashboard.
Endpoints at a glance
Three problems it solves best.
Problem. A product complaint goes viral on TikTok before the brand team sees it.
Outcome. Video-transcript indexing and a virality score surface the post within 90 seconds of inflection.
Problem. Equity-relevant chatter is scattered across Reddit, X and Discord.
Outcome. Ticker-tagged post streams with sentiment and virality feed directly into alpha signals and risk dashboards.
Problem. Identifying an emerging trend before it hits Google Trends takes hours of manual scrolling.
Outcome. Cross-platform topic clustering flags trends while they are still sub-viral, with drill-down to source posts.
Why teams pick FreshGeo.
- ✓ Video transcript indexing on TikTok and YouTube, not just captions
- ✓ Virality scoring across platforms, not per-platform volume charts
- ✓ Real-time streaming via websocket, not 15-minute dashboard polls
- ✓ Transparent pricing per million posts, not seat licences
- ✓ Entity and sentiment tagging included, not a separate NLP bolt-on
- ✓ Historical backfill up to 24 months on request
- ✓ API-first with no mandatory UI; build your own workflow
"We caught a product-quality issue trending on TikTok four hours before it hit Twitter. That window saved a launch."
- ✓ Real-time data
- ✓ Advanced sentiment
- ✓ Webhooks
- ✓ Priority support
- ✓ Everything in Pro
- ✓ Custom feeds
- ✓ Historical data
- ✓ Dedicated support
Questions, answered.
Which platforms are covered? +
Reddit, TikTok, YouTube, X, Threads, Instagram (public), Bluesky, Mastodon, Discord public servers, Twitch chat, Trustpilot, App Store and Play Store reviews, plus about 30 vertical forums. New platforms are added on customer request when public APIs or ToS allow.
Do you transcribe TikTok and YouTube video? +
Yes. Priority queries trigger on-demand ASR within 60-90 seconds of publication for TikTok and YouTube. Transcripts are searchable as text, with entity extraction and sentiment applied to spoken content as well as written captions.
How does virality scoring work? +
The virality score is a 0-100 index blending post velocity, estimated reach, cross-platform spread and engagement rate versus the author's baseline. It is designed to detect inflections early — a score crossing 60 is usually 2-6 hours ahead of mainstream trend trackers.
Is the sentiment model multilingual? +
Yes. Sentiment and entity extraction work across English, French, German, Spanish, Italian, Portuguese, Dutch, Polish and Nordic languages natively, with a fallback multilingual model for 40+ others. Calibration scores per language are published in the docs.
How is this different from Brandwatch or Meltwater? +
FreshGeo is API-first with no mandatory dashboard, prices per million posts rather than per seat, and includes video transcript indexing at no extra cost. Coverage leans harder into TikTok, Discord and niche forums where consumer trends now start, rather than legacy news and blogs.
Can I get historical data? +
Yes. Backfill is available up to 24 months for most platforms and 12 months for TikTok and Discord. Historical pulls are priced per million posts matched and delivered as Parquet or JSONL to S3, GCS or Azure Blob within 24 hours of request.
Ship your first request in 4 minutes.
Free tier. No card required. UK-hosted, UK-supported.