عربيDownload
FeaturesExtensionDocsChangelogعربي
Install NishioGetting started · Nishio is a desktop app for macOS, Windows and Linux. Download the installer for your system, open it once, anFirst run: your local accountGetting started · The first time you open Nishio you create a private owner account and your first workspace. The account existsAdd your YouTube API key (optional)Getting started · With your own free YouTube Data API key, Nishio can run official searches and fetch channel uploads. Without oDesktop basicsGetting started · How the window, menus and shortcuts work, where your data lives, and how to update without losing anything.Start a channelResearch · Turn a channel idea into a plan in three steps: describe the idea, study real channels in a saved sample, thenResearch inbox & today’s opportunitiesResearch · The Research inbox shows “Today’s opportunities”: niches found by a daily official search, each scored only whDiscoverResearch · Discover searches everything you have saved in this workspace and lets you narrow it down with research filterNiche finder catalogResearch · A built-in editorial catalog of well-known YouTube niches, grouped by category, format, production style and r
DocsResearch

Thumbnails

See saved videos as a thumbnail wall with their numbers, and study images you are allowed to process with an optional local vision model.

Applies to: Desktop 0.1 · Extension 0.1

Real screenshot of the appScreenshot of Thumbnails

The thumbnail wall

Thumbnails uses the same research filters as Discover, but shows each saved video with its thumbnail, title, channel, date, Views, Outlier ratio and Comparable sample. “Inspect evidence” opens the source; Save adds it to your Swipefile.

Authorized image research

  1. Scroll to “Authorized image research”.
  2. Enter a Title and the Source URL, and choose a PNG, JPEG or WebP image — maximum 3 MB.
  3. Tick “I own this image or have permission to process it locally.” and import it.
  4. Optionally use “Analyze with local vision model” (needs a model assigned to Image analyst in Model manager) or “Find visually similar saved images”.

Visual similarity uses a perceptual hash distance: it finds images that look alike, not images with a similar meaning.

One image proves littleAnalysis describes one uploaded image. It cannot establish whether a channel is faceless, its CTR, or why a video succeeded. Faceless and AI-content signals stay unknown without visual evidence.