# Video Lab Idea Engine

> Find proven video ideas through your AI assistant. Search 170M+ YouTube videos and Shorts from curated channels, refreshed daily, ranked by how far each video outperformed its own channel.

Section: NexLev MCP  
Badge: New  
Canonical page: /docs/mcp/video-lab-idea-engine

---

Find video ideas that are **already proven to work**. The Video Lab Idea Engine searches **170M+ YouTube videos and Shorts** from a curated list of **quality channels** and ranks them by how much each video **outperformed its own channel**. A video that pulled 10× its channel's usual views tells you the topic did the work, and that is an idea you can make your own.

It works for every kind of creator:

- **Faceless channels**: AI voice, stock footage, animation, compilations.
- **On-camera creators**: vloggers, educators, reviewers, gamers.
- **Long-form and Shorts**, in any niche and any language.

It uses the same data as the **Video Lab** page in the NexLev app, so what your assistant finds matches what you see there.

## Prerequisites

- **Connected AI assistant**: Claude, ChatGPT, or any MCP-compatible client. See the [Quickstart](/docs/mcp/quickstart).
- **A paid NexLev plan**: Lite, Premium, or Pro. Free accounts cannot use the Idea Engine. [Upgrade your plan](https://dashboard.nexlev.io/?required_nexlev_lite_pro_access=true).

> **Note:** Your AI assistant calls this tool for you. Describe the ideas you want in
> plain language, and it picks the filters, runs several searches from different
> angles, and explains why each idea worked.

## What It's For

### Find ideas that are working right now

Ask for a niche and a time window, and get the videos that beat their channel's average by the widest margin.

### Copy the topic, not the channel size

A high outlier score on a **small channel** is the most repeatable kind of idea: the channel had no big audience to lean on, so the topic and packaging drove the views.

### Study title patterns before you write

Pull the top outliers for a topic and have your assistant point out the title formulas, hooks, and angles they share.

### Research Shorts separately

Switch to Shorts to see which short-form topics are breaking out, with their own ranking and filters.

## Understanding the Outlier Score

**Outlier score = the video's views ÷ the channel's average views per upload.**

| Score   | What it means            |
| ------- | ------------------------ |
| **1**   | Average for that channel |
| **2+**  | Strong                   |
| **5+**  | Exceptional              |
| **10+** | Breakout                 |

> **Tip:** For idea research, start with a **minimum outlier score of 3** and a
> **subscriber cap** (for example, under 100K). That surfaces ideas that worked
> because of the topic, not because of the channel's size.

## When to Use It

| You want…                                    | Use                                                                              |
| -------------------------------------------- | -------------------------------------------------------------------------------- |
| **Video ideas** in a niche, topic, or format | **Video Lab Idea Engine** (this page)                                            |
| **Channels** to study or compete with        | [`search_niche_finder_channels`](/docs/mcp/channel-search)                       |
| One channel's own videos or outliers         | [`youtube_channel_videos` / `youtube_channel_outliers`](/docs/mcp/youtube-tools) |
| Videos similar to **one specific video**     | [`get_similar_videos`](/docs/mcp/similar-videos)                                 |

> **Note:** The Idea Engine **replaces `search_videos`**, which searches an older index.
> `search_videos` still works for existing setups, but new research should use
> the Idea Engine.

## Tool

### `video_lab_idea_engine`

Searches YouTube videos or Shorts by title keywords and filters, and returns them with their outlier score and channel details. It can also list which filter values are available for a search, so your assistant can pick exact categories, languages, or countries.

**Best for:**

- "Give me video ideas about personal finance that are working right now"
- "Find outlier videos on small channels in the true crime niche"
- "What faceless history videos went viral in the last 30 days?"
- "Show me Shorts ideas about cooking hacks"
- "Title patterns of breakout Minecraft videos"

**What it returns:**

- **Video**: title, link, thumbnail, views, outlier score, upload date (long-form), and length (long-form)
- **Channel**: name, handle, subscribers, average views per upload, category, language, country, quality tier, and whether it is faceless
- **Totals and paging**: how many videos matched, and how to get the next page

**Example prompts:**

```
Give me 10 video ideas about personal finance that are working right now. Only
videos from the last 30 days with an outlier score of 3 or more, on channels
under 100K subscribers. Skip news and music channels, and tell me why each one
worked.
```

```
Find faceless history videos in English with an outlier score above 5. Show
the smallest channels first and summarize the title patterns they share.
```

```
Show me YouTube Shorts ideas about cooking hacks with an outlier score of 3 or
more.
```

```
What are the evergreen breakout videos in the true crime niche? Use channels'
all-time top videos, not just recent uploads.
```

```
Find on-camera tech review videos between 8 and 20 minutes long that got at
least 5x their channel's average views this year.
```

### Modes and formats

| Parameter | Values                       | What it does                                                                                                                                                |
| --------- | ---------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------- |
| `mode`    | `search` (default), `facets` | `search` returns videos. `facets` returns the filter values available for the same search, with counts (categories, languages, countries, tiers, and more). |
| `format`  | `long` (default), `shorts`   | Long-form videos or YouTube Shorts. Shorts have no upload date or length (see [Shorts](#shorts)).                                                           |

### Filters & parameters

All parameters are optional. List parameters take comma-separated values, for example `en,es`.

| Parameter                             | Type                 | Description                                                                                                                                             |
| ------------------------------------- | -------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------- |
| `query`                               | string               | Title keywords, matched as a phrase and as words, with tolerance for typos. `*` (default) browses by filters only. Up to 200 characters.                |
| `isExactMatch`                        | `"true"` / `"false"` | `"true"` keeps only titles that contain the exact phrase.                                                                                               |
| `category`                            | list                 | Only these channel categories (see [category keys](#category-keys)).                                                                                    |
| `excludeCategory`                     | list                 | Hide these categories. `news,music` is a good default, since those dominate many keyword searches.                                                      |
| `faceless`                            | boolean              | Leave out for all creators. `true` = faceless channels only. `false` = on-camera channels only.                                                         |
| `language`                            | list                 | Channel language codes, e.g. `en,es`.                                                                                                                   |
| `country`                             | list                 | Channel countries as full names, e.g. `United States,Canada`.                                                                                           |
| `tier`                                | list                 | Channel quality, `A,B,C` (A = best). Not every channel has a tier, so this narrows results a lot.                                                       |
| `audience`                            | list                 | `western`, `south_asian`, `other_non_western`, `unclear`.                                                                                               |
| `source`                              | `latest` / `popular` | `latest` = channels' recent uploads (about the last 3 months). `popular` = channels' all-time top videos, good for evergreen ideas. Leave out for both. |
| `minOutlierScore` / `maxOutlierScore` | number               | Outlier score range. 3+ is recommended for idea research.                                                                                               |
| `minView` / `maxView`                 | number               | Video view range.                                                                                                                                       |
| `minSubCount` / `maxSubCount`         | number               | Channel subscriber range.                                                                                                                               |
| `minLength` / `maxLength`             | number or `"mm:ss"`  | Video length in seconds or `mm:ss`. Long-form only.                                                                                                     |
| `startDate` / `endDate`               | `YYYY-MM-DD`         | Long-form: uploaded on or after / on or before. Shorts: first seen by NexLev on or after / on or before.                                                |
| `excludedKeywords`                    | list                 | Drop videos with any of these words in the title, e.g. `lyrics,official video`.                                                                         |
| `minChannelBaselineViews`             | number               | Minimum channel average views per upload. Default **100**, which hides ad-driven channels whose uploads get only a few dozen views. `0` turns it off.   |
| `channelId`                           | string               | Only this channel's videos (`UC…` ID).                                                                                                                  |
| `sortBy`                              | string               | One or two sort fields, comma-separated (see [sorting](#sorting)). With a query and no `sortBy`, results are ranked by how well the title matches.      |
| `sortOrder`                           | string               | `asc` or `desc` for each sort field, e.g. `desc,asc`.                                                                                                   |
| `limit`                               | number               | Results per page, 1–100. Default **20**.                                                                                                                |
| `page`                                | number               | Page number, for the first 10,000 results.                                                                                                              |
| `searchAfter`                         | array                | Cursor from the previous response, for results past 10,000. Pass it back unchanged.                                                                     |

### Category keys

`music`, `gaming`, `entertainment_commentary`, `lifestyle_vlog`, `education`, `sports`, `religion`, `automotive`, `diy_home_crafts`, `compilation_reupload`, `travel`, `finance_business`, `health_fitness`, `other`, `tech_reviews`, `comedy`, `kids`, `self_improvement`, `documentary_storytelling`, `art_design`, `science_tech`, `food_cooking`, `film_tv_review`, `beauty_fashion`, `asmr_sleep_ambient`, `politics_commentary`, `nature_animals`, `news`, `history`, `true_crime_mystery`

Not sure which value to use? Ask your assistant to run a **facets** search first, and it will list the categories, languages, and countries that actually have results for your topic.

### Sorting

| Format    | Sort fields                                                                                                   |
| --------- | ------------------------------------------------------------------------------------------------------------- |
| Long-form | `outlierScore`, `viewCount`, `publishedAt`, `firstSeenAt`, `channelSubCount`, `lengthSeconds`, `channelScore` |
| Shorts    | `outlierScore`, `viewCount`, `firstSeenAt`, `latestRank`, `channelSubCount`, `channelScore`                   |

Use two fields to break ties. For example, `sortBy: "outlierScore,channelSubCount"` with `sortOrder: "desc,asc"` puts the biggest outliers first and, among equal scores, the smallest channels first.

### Example calls

**Breakout ideas on small channels, last 30 days:**

```json
{
  "query": "budgeting",
  "minOutlierScore": 3,
  "maxSubCount": 100000,
  "startDate": "2026-10-21",
  "excludeCategory": "news,music",
  "sortBy": "outlierScore",
  "sortOrder": "desc"
}
```

**Faceless history videos in English, best first:**

```json
{
  "query": "*",
  "category": "history",
  "faceless": true,
  "language": "en",
  "minOutlierScore": 5,
  "sortBy": "outlierScore,channelSubCount",
  "sortOrder": "desc,asc"
}
```

**Shorts ideas:**

```json
{ "format": "shorts", "query": "cooking hack", "minOutlierScore": 3 }
```

**See which filter values exist for a topic:**

```json
{ "mode": "facets", "query": "minecraft" }
```

### Response

The assistant gets a one-line summary, such as `Found 10000+ videos; showing 20.`, plus the full data:

```json
{
  "format": "long",
  "query": "budgeting",
  "total": 10000,
  "totalRelation": "gte",
  "videos": [
    {
      "videoId": "abc123def45",
      "url": "https://www.youtube.com/watch?v=abc123def45",
      "title": "I Tried Living on $1,000 a Month",
      "views": 1250000,
      "outlierScore": 14.2,
      "publishedAt": "2026-11-02T00:00:00.000Z",
      "published": "2 weeks ago",
      "lengthSeconds": 734,
      "thumbnail": "https://i.ytimg.com/vi/abc123def45/hqdefault.jpg",
      "source": ["latest"],
      "channel": {
        "id": "UCxxxxxxxxxxxxxxxxxxxxxx",
        "title": "Example Channel",
        "handle": "@example",
        "subscribers": 48000,
        "avgViewsPerUpload": 88000,
        "category": "finance_business",
        "language": "en",
        "country": "United States",
        "tier": "A",
        "faceless": true
      }
    }
  ],
  "searchAfter": [1234, "abc123def45"],
  "pagination": { "currentPage": 1, "totalPages": 417, "limit": 20 },
  "appliedFilters": { "query": "budgeting", "minOutlierScore": "3" }
}
```

- `total` is exact up to 10,000. `totalRelation: "gte"` means **10,000 or more**.
- Shorts results have `firstSeenAt` instead of `publishedAt`, `published`, and `lengthSeconds`, and their `url` is `https://www.youtube.com/shorts/<id>`.
- A `facets` search returns the filter values with counts instead of videos:

```json
{
  "format": "long",
  "facets": {
    "total": 48210,
    "category": [{ "value": "gaming", "count": 1234 }],
    "language": [{ "value": "en", "count": 30122 }],
    "country": [{ "value": "United States", "count": 15870 }],
    "audience": [{ "value": "western", "count": 26411 }],
    "tier": [{ "value": "A", "count": 4120 }],
    "source": [{ "value": "latest", "count": 21004 }]
  }
}
```

**Pagination:** ask your assistant to "show more" or "get the next page". It handles `page` for the first 10,000 results and passes `searchAfter` back for anything past that.

## Shorts

Set `format` to `shorts` (or just ask for "Shorts ideas") to search YouTube Shorts.

- **No upload date or length.** YouTube does not give Shorts a publish date, so `startDate` and `endDate` filter the date **NexLev first saw the Short** instead. Length filters do not apply.
- **Different sort fields.** Shorts can also sort by `latestRank`; they cannot sort by `publishedAt` or `lengthSeconds`.

## Data Freshness

- **Curated channels**, covering 170M+ videos and Shorts.
- **Long-form videos** refresh every **8 hours** (gaming channels more often).
- **Shorts** refresh **daily**.
- The **Video Lab** page in the NexLev app uses the same index, so results match what you see there.

## Real-World Workflows

These prompts chain the Idea Engine with other NexLev tools.

### Turn an outlier into your next video

```
Find the top 5 outlier videos about "budget travel" from the last 60 days on
channels under 50K subscribers. Pick the one with the highest score, get its
transcript, and write me an outline for my own take on the same idea with a
fresh angle.
```

### Build a title swipefile

```
Search for breakout videos about "stoicism" with an outlier score of 5 or
more. List the 15 best titles, group them into title formulas, then save the
top 5 videos to my "Title Ideas" swipefile folder with the formula as the
note.
```

### Compare faceless and on-camera ideas

```
For the topic "home workouts", run one search for faceless channels only and
one for on-camera channels only, both with a minimum outlier score of 3. Which
ideas work in both, and which only work with a face on camera?
```

### Study the thumbnails behind the ideas

```
Find the 10 biggest outlier videos about "AI tools" this month, then find
similar thumbnails for the top 3 and tell me what the winning thumbnails have
in common.
```

### Explore a niche in another language

```
Run a facets search for "receitas fáceis" to see which languages and countries
have results, then show me the top Portuguese-language outliers from Brazil.
```

Your assistant brings in the other tools on its own: [transcripts](/docs/mcp/video-content#transcript) to study a video, [Similar Thumbnails](/docs/mcp/similar-thumbnails) for packaging, [Channel Analytics](/docs/mcp/channel-analytics) for the channel behind an idea, and [Swipefile](/docs/mcp/swipefile) to save what you find.

## Allowance and Pricing

| Plan    | Per hour | Per day | Per week |
| ------- | -------- | ------- | -------- |
| Free    | —        | —       | —        |
| Lite    | 60       | 150     | 600      |
| Premium | 60       | 150     | 600      |
| Pro     | 120      | 300     | 1,200    |

- Reaching **any** of the hourly, daily, or weekly numbers stops included use until that window resets.
- Calls also count toward your account's [weekly MCP budget](/docs/mcp/quickstart#weekly-limit).

### Past your allowance

With **usage credits** turned on, Lite, Premium, and Pro plans can keep going at **$0.05 per call**, up to **1,000 calls a day** in total. Without usage credits, the tool stops at your limit and tells you when it resets.

> **Tip:** **Run several narrow searches instead of one broad one.** Try synonyms,
> sub-topics, formats, and languages. Each search is quick, and different angles
> surface ideas a single search misses.

## Troubleshooting

**Too few results**

- Lower `minOutlierScore`, widen the date range, or remove `tier`. Not every channel has a tier, so it narrows results a lot.
- Turn off `isExactMatch` or shorten the query to fewer words.
- Set `minChannelBaselineViews` to `0` to include very small channels.

**Results full of music or news**

- Add `excludeCategory: "news,music"`, or drop title words like `lyrics,official video` with `excludedKeywords`.

**Shorts date filter seems off**

- For Shorts, dates are when NexLev **first saw** the Short, not when it was uploaded. A Short can be older than its first-seen date.

**Total says 10,000**

- `totalRelation: "gte"` means there are 10,000 **or more** matches. Add filters to narrow the search, or page past 10,000 with `searchAfter`.

**Tool not available**

- The Idea Engine needs a Lite, Premium, or Pro plan. [Upgrade your plan](https://dashboard.nexlev.io/?required_nexlev_lite_pro_access=true).

## FAQ

### How is this different from `search_videos`?

`search_videos` searches an older video index and is deprecated. The Idea Engine searches a newer, larger index of curated channels, refreshed daily, ranks by outlier score, and adds Shorts, faceless filters, and facets. Existing setups that call `search_videos` keep working.

### Is it only for faceless channels?

No. By default it covers every creator. Set `faceless` to `true` for faceless channels only, or `false` for on-camera channels only.

### Can it find channels?

No. It returns videos. To find channels, use [Channel Search](/docs/mcp/channel-search).

### Why do small channels with high scores matter most?

A big channel gets views from its audience no matter the topic. When a small channel gets many times its usual views, the idea itself drove the results, so it is much more likely to work for you too.

### Does it match the Video Lab page in the app?

Yes. Both use the same index, so results line up with what you see in the app.

## What's Next?

- **[Trending Videos](/docs/mcp/trending-videos)**: What YouTube is suggesting right now, and viral hits from small channels
- **[Similar Videos](/docs/mcp/similar-videos)**: Find more videos like an idea you picked
- **[Video Content](/docs/mcp/video-content)**: Transcripts to study how a winning video is structured
- **[Video Frame Extraction](/docs/mcp/video-frame-extraction)**: Break down how an outlier is edited, frame by frame
- **[Swipefile](/docs/mcp/swipefile)**: Save the ideas worth making
