Hermes Agent Video Analysis: A Step-by-Step Guide for Small Businesses

Hermes Agent video analysis lets a small business ask an AI agent questions about its own videos, from a ten-second product clip to a long training recording, and get written answers, transcripts and timestamps back. There are two ways to do it: a tool built into Hermes, and a new community plugin called video-xray that joined the plugin catalog on 15 November 2026. Here’s how a business owner sets both up, one step at a time, and what to check before staff footage goes anywhere.

What Hermes Agent video analysis does for a small business

Most small businesses sit on video they never fully use: product clips, recorded demos, staff training, webinars and customer clips sent over chat. Hermes can now read that footage for you. According to the Hermes tools reference, the built-in tool covers captions, scene breakdowns, key timestamps and visual descriptions.

Neither one makes new videos. If that’s what you’re after, read the guide to Hermes Agent AI video generation instead.

Step 1: Pick the videos worth analysing

Start with footage your team already reviews by hand, because that’s where an agent can take work off someone’s plate.

• Product and social clips — check on-screen text, labels and which product shows first.
• Training videos — turn them into subtitles and a written checklist.
• Recorded webinars — pull out quotes with timestamps for a recap.
• Ads — break a finished ad into shots for your next brief.

Leave out anything you don’t have permission to share with an AI provider. Customer and staff footage needs consent first.

Step 2: Switch on the video tool in Hermes

The video toolset is opt-in, so it isn’t loaded by default. The toolsets reference says to add it explicitly. Either start Hermes with the flag, or list it in your config:

hermes --toolsets video

# or in ~/.hermes/config.yaml
toolsets:
  - video

Next, give Hermes a model that can actually watch video. The Hermes source code sends every clip to an auxiliary model, never straight to your main chat model, and names Gemini as an example. Set the provider and model under auxiliary.vision in the configuration docs. If no auxiliary vision client is set up, the tool simply won’t appear.

Step 3: Ask your first question

Pick a short clip saved on your computer and ask something specific. For example: “Use video_analyze on shop-promo.mp4. Is the price label readable, and when does the logo first appear?”

• It reads mp4, webm, mov, avi, mkv and mpeg files.
• Anything above 20 MB can take a while.
• Files are capped at roughly 50 MB, so cut long recordings down first.
• Public video links work too, unless the site is on your Hermes blocklist.

Hermes tells the model to describe motion, audio cues, text overlays and scene changes before answering, so even a narrow question comes back with useful context.

Step 4: Add video-xray for long recordings

The 15 November 2026 canary release added video-xray to the Hermes plugin catalog. Its catalog entry lists version 0.2.2, Hermes 0.22.1 or newer, community tier and Windows as the only verified platform. The plugin’s README sets it up in three moves:

• Copy the plugin files into the Hermes plugins folder under your Windows local app data.
• Enable it with hermes plugins enable video-xray and restart Hermes Desktop completely.
• Copy the desktop panel file into the desktop-plugins folder and press Ctrl+Shift+R.

You’ll also need ffmpeg installed. After that, a gold workbench strip appears above the chat box. Pick a time window, choose 8 to 64 frames per segment, choose a speech engine (Whisper is the suggested start), and run the job. You get a Markdown report, SRT subtitles for CapCut or Premiere, and a gallery of every frame it looked at.

Step 5: Set privacy rules before staff use it

The catalog requires a security note on every plugin, and video-xray’s is short and clear. Read it before anyone attaches a real file:

• Every attached video starts an analysis job when you press send, unless you untick Analyze attached videos on send.
• Video frames go to your Hermes chat model, so a cloud model means a cloud provider sees them.
• Audio goes to a cloud speech service only if you set one.
• A saved cloud speech key is stored in plain text in an owner-only file.

For anything private, use a local chat model and keep auto-analysis off. Videos customers send in Telegram are also covered: the Telegram docs say a posted clip is cached and only opened with video_analyze when someone asks about it.

My verdict

For most small businesses, the built-in tool is enough to start. Turn on the video toolset, point it at a video-capable model, and test it on five clips your team already reviews. Add video-xray only if you run Hermes Desktop on Windows and need transcripts and full reports from long recordings — and switch off auto-analysis on day one. Set up that way, Hermes Agent video analysis turns footage you already own into notes your team can actually use.

About Julian Goldie

I’m Julian Goldie — founder of Goldie Agency and the AI Profit Boardroom. I test new AI agents almost every day for 432K+ YouTube subscribers and run Hermes, OpenClaw, OpenCode, Claude and Codex side by side in my own Agent OS. Claim your free strategy session or join the AI Profit Boardroom.

Want another angle? Read the AI Profit Boardroom guide to Hermes Agent video analysis.

FAQ: hermes agent video analysis

Can Hermes Agent analyse a video file?

Yes. Once you enable the opt-in video toolset, video_analyze can answer questions about a clip from a link or a file on your computer.

Which AI model watches the video?

The built-in tool uses a separate auxiliary video-capable model, such as Gemini. The video-xray plugin uses the chat model Hermes Desktop is already running.

Is video-xray made by Nous Research?

No. It is a community plugin maintained by LuuMoss. Nous lists it in the Hermes plugin catalog at a reviewed, pinned version.

Does video-xray work on a Mac?

Its catalog entry lists Windows as the only verified platform, and the install steps use Windows folders.

How large can a video be?

Roughly 50 MB for the built-in tool, and it gets slower above 20 MB. video-xray lets you analyse one time window at a time.

Can it make videos for my business?

No. Both tools read video. Making new clips is a separate Hermes feature covered in the video generation guide.

Related reading

• Best Hermes Agent Plugins
• Hermes Agent Browser
• Hermes Delegation

Last updated November 2026. This page is a living guide to hermes agent video analysis — the facts here move fast and it is updated as they do.