Faceless / AI video intelligence
Find breakout niches ranked by measured view velocity, not by subscriber counts.
A YouTube niche finder that works backwards from evidence: instead of guessing which niche to start, it tracks 2557 videos that already outran their own subscriber base, then ranks the niches, formats and language markets those videos came from. It is built for faceless and AI-assisted channels, where the production cost is low enough that a working format gets reproduced quickly — so the useful question is what already has evidence behind it, not what sounds promising. Every signal below is public API data: view count, subscriber count, publish date and duration, plus a rule-based read of the hook, pacing and production choices. Metrics are captured snapshots refreshed when the dataset is re-scraped, not live counters — ages on this page are recalculated from each video's publish date as of the last rebuild (2026-09-25).
100.2×
90s Village Curd Rice 🍚 | Ghibli Style Village Life 🌾✨ | Ghibli inspired countryside
31.0M views from a 52.5K-subscriber channel — 590.4× its audience, averaging 326.3K/dayover 95 days. Published 2026-06-22.
Dreamy Village Life — channel breakdown →
Filter, sort or compare the full dataset — or read the static boards below. Rows are ranked by the breakout score: views per day divided by audience size, weighted by engagement; 1.0× is the dataset median.
A niche finder is only as good as the number it ranks by, so this section is the one that says where every figure on the page comes from. A video enters the dataset when it clears four conditions: at least 20.0K views, no more than 100.0K subscribers at capture, reach of at least 3× the channel's own subscriber count, and a publish date inside a rolling 180-day window. Those are entry conditions, not a ranking: a position on any board here is decided by the breakout score — views per day divided by the audience the channel started with, weighted by whether engagement looks normal for that niche. Every row prints the two components next to the score, so you can ignore the composite and argue with the parts.
What a YouTube niche finder of this shape cannot do is worth stating before you use it. Each video carries a single capture snapshot, so views per day is an average since publication rather than a current reading — nothing here can tell you whether a video is still climbing or already peaked. The score is also normalised against this dataset's own median, which means every multiple on this page is relative to the 2557 tracked videos and is not comparable with any other tool's numbers. And the sample is filtered in a specific direction: it contains low-subscriber channels that broke out, and nothing about the uploads in the same niches that went nowhere.
All 16 niches this YouTube niche finder tracks, ordered by how fast a typical tracked breakout in each one accumulates views. Speed alone is only half the answer, so the table carries both sides next to each other: median views per day is the "is anything actually moving here" question, and breakouts tracked is the "is everyone already here" question. The rows worth your time are the ones where the first is high and the second is still thin — that combination is what an unserved niche looks like in practice, and no opinion-based niche list can show it to you.
| # | YouTube niche | Breakouts | Channels | Median views/day | Median views/sub | Median length | AI-generated | Top market |
|---|---|---|---|---|---|---|---|---|
| 01 | Romance / Mini-Drama | 250 | 127 | 5.7K/day | 12.0× | 61m | 13% | English / Global (172) |
| 02 | Feature Films & Series | 115 | 68 | 5.7K/day | 15.7× | 60m | 22% | English / Global (92) |
| 03 | Builds / Transformations | 95 | 33 | 4.3K/day | 17.2× | 3m | 4% | English / Global (87) |
| 04 | Music / Performance | 159 | 47 | 4.3K/day | 11.6× | 4m | 17% | English / Global (136) |
| 05 | Short Films & Stories | 211 | 124 | 4.0K/day | 17.4× | 5m | 37% | English / Global (135) |
| 06 | Explainer / Science | 220 | 72 | 3.9K/day | 10.9× | 18m | 13% | English / Global (179) |
| 07 | Horror / Thriller | 132 | 72 | 3.6K/day | 14.0× | 20m | 11% | English / Global (103) |
| 08 | ASMR / Cozy | 290 | 93 | 3.4K/day | 20.6× | 6m | 34% | English / Global (253) |
| 09 | Documentary / History | 387 | 203 | 3.3K/day | 14.5× | 27m | 18% | English / Global (269) |
| 10 | Sci-Fi / Time Travel | 120 | 58 | 3.1K/day | 13.1× | 14m | 23% | English / Global (111) |
| 11 | Gaming / Anime | 70 | 37 | 2.8K/day | 12.2× | 10m | 30% | English / Global (67) |
| 12 | Fantasy / Myth | 116 | 63 | 2.8K/day | 16.5× | 19m | 17% | English / Global (96) |
| 13 | Growth / Monetization | 50 | 33 | 2.7K/day | 8.6× | 16m | 30% | English / Global (42) |
| 14 | Apocalypse / Survival | 75 | 26 | 2.1K/day | 11.3× | 11m | 0% | English / Global (75) |
| 15 | Kids / Learning | 106 | 43 | 1.9K/day | 20.0× | 11m | 20% | English / Global (90) |
| 16 | Tutorials / Tools | 161 | 102 | 1.4K/day | 9.8× | 10m | 42% | English / Global (112) |
"AI-generated" is the share of a niche's tracked breakouts whose own title declares AI-generated visuals — a rule-based inference from the title line, not frame-level analysis. Labels this site puts on an entry are not evidence: they are written by the same pipeline, so counting them would be circular. Every tracked breakout is a faceless format, so that marker is not a column: it is constant across the dataset and would sort nothing. Source: rule-based inference from title + duration + market (hookType / openingSeconds / structure / repeatableChoices); provenance: api metrics + metadata only, no frame-level video analysis.
Two limits to carry into the table. It cannot measure a niche's true size, because the dataset only contains videos that cleared the 20.0K-view entry filter — a niche full of 5,000-view uploads looks empty here even if it is crowded. And median length is the number most people skip and the one that decides production cost: the median upload across these niches runs from 3 to 61 minutes, which is the difference between a four-minute shorts niche and a documentary niche.
Market here means the language a video is actually spoken in, not the country of the channel: English / Global holds 2019 of the 2557 tracked breakouts (79%), and the remaining 538 split across 17 other markets — Hindi (213), Spanish (89), Portuguese (69), Arabic (75), Indonesian (10) and others. That distribution is not a finding about demand: this YouTube niche finder decides which markets get searched by its own keyword pool, so a language looks small here partly because it was searched for less. What the slice is good for is comparing how the same niche is played in two markets — the interactive filter above does exactly that, and the cross-market comparison pages lay it out side by side.
All 18 markets side by side → · Example: Romance / Mini-Drama across markets →
The twelve videos currently sitting at the top of the board, in the order the score puts them — one row per video, with its channel, niche, view count, daily rate and the reach multiple behind its position. "Latest snapshot" is meant literally: the ranking is recomputed from the captured metrics on every rebuild, while the view counts themselves were captured on the dates published on the breakout board. Rank is not a verdict on quality — it is a measure of which of these a small channel managed to push out to a far bigger audience than it started with.
Periods Ke Dard Mein Bhi Ghar Sambhalti Ek Maa ❤️ | Emotional Indian Village Story | Ghibli Inspired
Building a Tree Hut to Protect Peanut Farm 🥜| Ghibli Village Life #villagelife #rainsounds #farmlife
The breakout board — the top 20 by score, plus every niche board →
Supply is the half of the niche question most lists skip, and here it is simply how many breakouts the radar has found in each niche. Documentary / History is the deepest with 387 tracked breakouts from 203 channels, and Growth / Monetization is the thinnest with 50. Deep is not automatically bad — it means the format is known, so an idea in that niche will be recognised on sight — and thin is not automatically good, since a niche can be thin because nobody has tried it or because it does not work. The useful reading is always the pair: Romance / Mini-Drama currently carries the highest median velocity (5.7K/day), while Tutorials / Tools sits at the bottom of the same ranking (1.4K/day), and that gap is where the search for an angle usually starts.
One thing this ordering deliberately does not claim: that a crowded niche is saturated, or that a thin one is unexploited. Crowding is counted from a filtered sample — videos that already cleared the entry conditions — so it measures how much visible activity a niche has, not how much competition you would face on upload day. Pair it with the velocity table above rather than reading it alone.
Six of the 16 niches get a board on this page — the ones with the deepest supply and the widest spread of formats — each showing its five strongest tracked breakouts and a link into the full niche board sorted by every condition. The remaining 10 niches are summarised with their counts right underneath: they are still tracked, still filterable above, and still linked here, they just do not need a full board on the front page repeating the same data.
68 channels · median 5.7K/day · median 15.7× its audience · median runtime 60 minutes · 22% declare AI-generated visuals · largest market English / Global
124 channels · median 4.0K/day · median 17.4× its audience · median runtime 5 minutes · 37% declare AI-generated visuals · largest market English / Global
127 channels · median 5.7K/day · median 12.0× its audience · median runtime 61 minutes · 13% declare AI-generated visuals · largest market English / Global
93 channels · median 3.4K/day · median 20.6× its audience · median runtime 6 minutes · 34% declare AI-generated visuals · largest market English / Global
203 channels · median 3.3K/day · median 14.5× its audience · median runtime 27 minutes · 18% declare AI-generated visuals · largest market English / Global
33 channels · median 4.3K/day · median 17.2× its audience · median runtime 3 minutes · 4% declare AI-generated visuals · largest market English / Global
Each of these niches has its own board and its own filters — they are one link away rather than one scroll away. Counts are tracked breakouts and the number after each name is that niche's median views per day.
The page is built to be read in a specific order, from the widest question to the narrowest, and the four steps below are that order. Skipping straight to the videos is the common mistake: the interesting part of a breakout is rarely the video itself, it is what the niche around it looks like.
One lane in this dataset behaves differently enough to pull out separately: films and short films whose own title declares AI-generated visuals. There are 188 such entries from 112 channels in the current capture, 40 of them running 30 minutes or longer, and only 12% name the generation tool they used — on this dataset, the tool is usually invisible. Two pages carry that lane: AI movies ranks the whole corpus by real view velocity and lists who is making them, and AI short films covers the short-form end, where almost all of the activity is. Both state the same limit in the same words: "declares AI" is what a video's own title says about itself, and no frame was inspected to confirm it.
The reason this lane is worth watching is the production cost, not the novelty. A format that can be produced without a camera, a host or a studio gets reproduced fast, so the window between "this works" and "everyone is doing it" is shorter than in most of YouTube — which makes velocity the useful signal and subscriber counts a lagging one. The rest of this dataset is the same argument in other genres: see faceless YouTube niches for the no-host formats and short-form breakouts for the wider short-video boards.
The questions below are the ones this dataset can actually answer, in the wording people search for. Where it cannot answer, the entry says so instead of rounding the gap off.
A tool that ranks YouTube niches by evidence instead of opinion. This one tracks 2557 videos from 954 small channels — every one of them a video that outran the subscriber base it was published to — then ranks the 16 niches those videos came from by how fast a typical breakout in each niche accumulates views. The output is a shortlist of niches worth investigating, not a verdict on which one will work for you.
Compare two numbers per niche and read them against each other: median views per day (is anything actually moving here) and tracked breakout count (is everyone already here). High velocity with thin supply is what an unserved niche looks like in this dataset. Documentary / History is the deepest niche tracked with 387 breakouts, while Growth / Monetization holds 50 — and one honest caveat: the dataset cannot measure a niche's true size, because it only contains videos that already cleared the 20,000-view entry filter.
A tracked video has at least 20,000 views, comes from a channel with no more than 100,000 subscribers at capture, reached at least 3× its channel's subscriber count, and was published inside a rolling 180-day window. Videos arrive two ways — keyword searches and direct pulls from channels already in the dataset — and those conditions are applied when an entry first enters the dataset, not re-applied every time the data is refreshed.
Not this one, and the reason is structural: the public YouTube Data API exposes view, like, comment, subscriber and duration metrics, and no revenue field at all. Views are the only currency this dataset can count, so every ranking here is about attention, not money. The channel value calculator shows what can and cannot be estimated from public numbers, with the formula written out.
Because the intake does exactly that. The tracker's keyword pool is built around faceless and AI-assisted channels, where production cost is low enough that a working format gets reproduced quickly — so the evidenced niches here skew that way. That is a filter on the sample, not a finding about YouTube: a niche full of on-camera creators is not missing because it is weak, it is missing because it was never searched for.
Five ways into the same dataset, each built for a different decision — studying how AI films are made, building a niche without a host, pricing a channel, or checking whether a subscriber-count tracker is the right tool at all.
Every tracked film whose own title declares AI-generated visuals, ranked by how fast it accumulated views against the channel it was published on — with the runtime, market and production markers behind each one.
The short-form end of the same corpus, where most of the activity is: what AI shorts actually run, which hooks repeat, and how fast they move compared with the feature-length entries.
Which niches the tracked breakouts run without an on-camera host, and the production choices that repeat in each one.
Estimate what a channel is worth from its subscriber and view numbers, with the formula and its assumptions written out instead of hidden.
What a subscriber-count tracker cannot tell you about a small channel, and what this radar does differently — including what it deliberately does not do.
Everything below is built from the same 2557 tracked signals, so the numbers agree across pages. The breakout board documents the ranking method and the capture dates in full.
Data: YouTube Data API v3 public metrics · 2557 tracked signals · 954 channels · page rebuilt 2026-09-25. Analysis source: rule-based inference from title + duration + market (hookType / openingSeconds / structure / repeatableChoices); provenance: api metrics + metadata only, no frame-level video analysis.