YouTube Viral & Trend Radar

Find YouTube Shorts and long-form videos by topic or channel, compare their public performance, and measure view growth across repeated runs. Export your research as CSV, Excel or JSON, or connect it to your own workflow. No YouTube API key or sign-in is required.

What you can research

• Content ideas and video performance in a specific niche.
• Shorts and long-form uploads from channels you follow.
• Changes in exact public view counters between observations.
• Views relative to a channel's available subscriber count.
• A video's lifetime average view rate relative to the median of its tracked channel videos.

Each run returns a bounded sample. Viral scores, content categories, detected language and title-hook signals are heuristics for research.

Start with a topic

{
  "discoveryModes": ["search"],
  "searchQueries": ["skincare routine", "makeup tutorial"],
  "formats": "short",
  "maxResults": 25,
  "maxAgeHours": 168,
  "saveHistory": true,
  "historyProfile": "beauty-research"
}

Use short, long or any for formats. To monitor channels, select ["channel"] and supply public channelUrls or @handles. The Channel Video Growth Monitor example starts with NASA; replace it with your channels.

Measure growth over time

The first observation establishes a baseline. Repeat the same history profile, sources, country and format at least minIntervalMinutes later (60 minutes by default). Two sufficiently spaced observations can measure view growth. Three can support acceleration or decline comparisons when the data meets the thresholds.

• viewsDelta: change in the exact public view counter.
• measuredViewsPerHour: that change divided by elapsed observation time.
• averageViewsPerHourSincePublished: lifetime average, separate from measured growth.
• reachMultiple: views divided by available subscriber count; rounded subscribers are flagged.
• velocityVsBaseline: lifetime average view rate divided by the channel median for its tracked sample.
• creatorBaselineVideos: total unique videos tracked for that channel; baselineWindow limits the newest observations used for the median.

States include BASELINE, NEW, INSUFFICIENT_INTERVAL, RISING, STABLE, ACCELERATING, DECLINING, STALE_BASELINE and COUNTER_CORRECTION. Acceleration and decline are conditional signals. DECLINING can mean slower growth while total views still increase.

Output and exports

The dataset includes available titles, channel links, video URLs, format, public counters, publication dates and precision, ranking signals, observations and metric warnings. Missing values remain empty. The overview displays date precision and whether subscribers are rounded.

The run also provides a summary and an HTML research dashboard. Hook signals analyze the title text; the Actor does not watch the video, transcribe speech or analyze its visual opening.

Pricing

$0.02 per run start + $0.005 per video returned. A run returning 25 videos costs $0.145 at the default 1 GB memory setting, before any applicable taxes. The start event is charged per GB of allocated memory, with a minimum of one event; allocating more memory increases the start charge. Each returned video is billed again when returned on a later monitoring run.

The start charge applies even when no matches are returned. Apify platform usage during the run is included in this event-based pricing. Optional paid proxies can add cost. Use output and charge limits to control spending. Actual results can be lower than the requested maximum because of filters, available candidates or collection limits.

Data limits

• Publication dates can be date-only. For these rows, hourly recency filters allow up to 24 additional hours of uncertainty. Check publishedAtPrecision; a proxy does not guarantee exact timestamps.
• Search localization is not creator or audience geography.
• Language and category filters use text heuristics and can misclassify content. Unknown language is kept.
• Channel baselines use videos observed by this history profile, not every upload on the channel.
• Regional Trending is best-effort and can be empty for anonymous sessions. Use niche searches or channel monitoring as the main discovery methods.
• Exact counters and metadata can be unavailable or change. Rounded estimates are flagged separately; missing comment counts or durations are not invented.
• Ranking describes observed public signals, not a prediction of future views, a probability of virality, or a guarantee of commercial performance.

The Actor requests public YouTube pages directly. It does not call another Actor, download media or sign in.