Live AI agents that can see

Build an AI agent that can see what your customers show it.

VoiceScholr is a free, no-code builder for live AI agents that see your customer’s camera or shared screen and talk them through the problem in real time. Share yours as a link, or put it on your website.

No card. No app or account for your customers. Calls aren’t recorded.

ReadyAI agent

VoiceScholr Guide (test)

Hi, I'm the VoiceScholr guide. I can see you, answer anything about VoiceScholr, and book a call with Abhishek. What would you like to know?

You’ll talk to an AI. It sees your camera or screen while you talk.
Free sessions last 3 minutes.

  • Hold it close, in good light
  • Ask “What do you see?”
  • Talk over it anytime
  • Ask “What does it cost?”
  • Hold something up and ask “What is this?”
  • Say “Book me a call with Abhishek.”

Free calls last up to 3 minutes. It’s an AI, so it can get things wrong.

A home Wi-Fi router on a bookshelf with one red light, seen through the customer's phone camera
Sees the camera · 1 frame a secondIllustration

169 calls this week · built on Stream Vision Agents and Gemini Live

Chatbots and phone lines can’t see the problem.

A dented bumper, a router’s lights, a settings page: people struggle to put it into words. Your agent looks at it with them instead.

The problem

Customers can't describe what they're looking at.

With VoiceScholr

So they show it. Your agent watches their camera or shared screen, about a frame a second, and talks them through it by voice.

The problem

“Please send us a photo” turns into a week of emails.

With VoiceScholr

Your agent looks at it with them, live. It asks for the angle it needs and takes a sharp snapshot to read a label or serial number.

The problem

Products come back “broken” when setup was the real problem.

With VoiceScholr

Put your agent's link on the box or your help page. It watches the setup and walks them through it, using the manual you give it.

The problem

Visitors leave when the only way to ask is “book a demo”.

With VoiceScholr

Add a Talk button to your site. Visitors get answers from your own facts straight away, with no form and no account.

Three ways to start.

Pick the one that fits today. You can do the others later.

Build it yourself

Free, with no card. Describe your agent in one sentence, or start from a template and edit it.

Start free
  1. 1

    Describe it, or pick a template

    One sentence is enough: we write the instructions. Paste your prices, policies or manual as its knowledge.

  2. 2

    Talk to it in your browser

    Test it with your camera or a shared screen before anyone else sees it.

  3. 3

    Share the link, embed it, or add a Talk button

    Send the link in an email or a text, print it as a QR code, or put the agent on your website.

Put it on your site

Share the link, paste the embed code into any page, or add a floating Talk button with one script tag. Shopify stores get a voice shopping assistant.

Talk to a person

Want to see it first, or ask a question before you build? Book a call with Abhishek through our own agent, or email us.

Built to see, not to show a face.

Your agent’s job is to look at your customer’s side of the call: the dent, the label, the settings page. A talking face is optional and off by default, so there’s no avatar to render or pay for.

Sees the camera or a shared screen

About one frame a second, while it talks in real time.

Answers from your facts

Paste your prices, policies or manual. It looks them up when a question needs them.

Your customer stays in control

Camera and mic only when they press Talk. Calls aren’t recorded.

What it doesn’t do.

The honest limits, so nothing surprises you after you’ve put it on your site.

  • Browser only. No phone numbers.
  • Free calls are capped per visitor (10 a day), and up to 6 can run at once across everyone.
  • About one frame a second: fine for a label or a screen, not fast movement.
  • No recordings, and VoiceScholr keeps no transcripts. If the business running the agent connects a webhook, it is sent the transcript when the call ends.
  • A snapshot photo, if the caller agrees to one, is kept 7 days, then deleted.
  • The optional face is a preset character, not a replica of a real person.
  • English only, for now.

Free to start, or bring your own key.

  • Free: 10 calls a day per visitor, 3 minutes each, on the included Gemini Live model.
  • Pro (coming soon): house minutes you pay for, no key needed.
  • Your own key: $0 to VoiceScholr at any volume; calls up to 10 minutes, no daily cap.
See pricing

Questions people ask about AI agents that can see.

Can an AI agent see what I show on my camera or screen and respond right away?

Yes. A VoiceScholr agent watches the camera or a shared screen about once a second while it talks with you in real-time voice, so it reacts to what it sees mid-conversation.

Do my customers need to download an app or create an account?

No. They open your agent's link in their phone or computer browser and press Talk. The browser asks for the camera and microphone only then, and they can start without the camera.

Is it free, and how long can a call last?

Yes, it's free to start, with no card. On the included Gemini Live model each visitor gets 10 free calls a day of up to 3 minutes. Add your own model key for calls of up to 10 minutes with no daily limit; you pay the model provider directly. See pricing

How do I add the agent to my website?

Three ways: share its link, paste the embed code into any page, or add a floating Talk button with one script tag. You can limit the embed to your own websites.

Can the agent book meetings?

Not yet for agents you build. Today only our own demo agent books calls, with Abhishek, the founder. Your agents answer, look and explain.

Are calls recorded? What happens to the video?

Calls are not recorded. The transcript is kept in server memory so the caller can see it, then deleted 10 minutes after the call. If the caller agrees to a photo, it is saved privately for the agent's owner and deleted after 7 days. The live stream goes to the AI model provider, Google Gemini by default, whose own terms apply. The default model uses Google's free tier, where Google may use call content to improve its products, so don't show anything sensitive.

Which AI models does it use?

Gemini 3.8 Live is included free. With your own API key you can switch an agent to Qwen Omni or OpenAI's realtime model, or to xAI Grok Voice, which is voice only. Keys are encrypted and never shown again.

What is an AI video agent, and does it need a face?

It's an AI you talk with over a live video call. Many tools focus on showing you an AI face; a VoiceScholr agent is built to see your side of the call, the camera or shared screen. A talking face is optional and off by default.

Your first agent is one sentence away.