ChatGPT Live camera access: use snapshots for visual help
A spoken conversation is more useful when your assistant can see the worksheet, project, or object you mean. If your current conversation has no live camera input, requested snapshots can provide visual context at the moments you choose. LetItSee supplies those images through MCP to a compatible assistant; voice comes from your AI client.
Can ChatGPT Live use your camera or share your screen?
Checked October 17, 2026: OpenAI's Voice documentation says Live does not support video or screen sharing. The previous Advanced experience still supports those features for eligible subscribers on iOS and Android. This is a difference between Voice modes, not the removal of every ChatGPT camera feature.
The same documentation says Live accepts images and can use plugins and connected apps available to your account. That does not establish that LetItSee is installed, accessible, or tested in your particular Live session. Check your current mode, app permissions, and available tools before choosing a workflow. The official reference is linked below.
Use a new image when the work changes
Ask for one photo after writing a new math step, finishing a physical test, or changing a model. Discuss that view, make your next change, and ask for another only when it will help. For something already on your computer, select a desktop window or tab instead of aiming a camera at the monitor.
For a single saved worksheet or existing screenshot, attaching it directly may be enough. A connected camera becomes useful when you keep changing the work in front of it and want the next requested view without repeatedly taking, transferring, and attaching photos yourself.
Check the tools in the conversation you will use
Your client needs a vision-capable assistant, remote MCP over Streamable HTTP, OAuth sign-in, and support for image results from tools. Connect LetItSee using your client's supported connection flow and sign in. Open the separate capture page with the same account, choose the camera or desktop share yourself, and confirm the preview.
Request one test image in the actual conversation and ask the assistant to describe an obvious detail. If the voice session cannot call the tool or use the returned image, continue in a compatible text conversation or attach a photo directly. Connecting an MCP server elsewhere does not automatically make it available in the current voice mode. Keep a phone capture page visible and the phone awake; switching apps can suspend its camera.
Keep the preview local and send useful checkpoints
LetItSee keeps the camera or screen preview on your device and uploads a JPEG only for a requested capture. Reuse the image already in the conversation while discussing the same state. Avoid repeated captures while you are thinking, explaining, or waiting for a task to finish.
This gives you control over which visual checkpoints are sent. The browser's camera can remain active for the local preview between requests; stop or close the capture page when finished. We have not measured cost, energy, or token savings against live video, and your AI provider's image and voice charges still apply.
Use the same visual workflow with other assistants
Voice availability and camera access are separate client capabilities. Claude's voice documentation, for example, describes connected tools and notes that not every result can be displayed in voice mode; it does not verify this camera workflow. Check image-tool support in the specific session rather than assuming a microphone button also gives the model a camera view.
An assistant with no voice mode can still use requested snapshots if it supports the MCP and image requirements above. Type the same request and discuss the result in text. LetItSee does not record microphone or system audio and does not supply continuous video or unattended observation.
Prompt to adapt
Take one fresh image of the [worksheet, object, or selected window] I have framed. Describe [specific detail] and help me decide the next step. Use this image while we discuss it. Wait for me to request another view after I change the work.
Source documentation
Set up without starting a capture
Set up your MCP client with the full tool reference. The separate capture page can replace the active capture page for your account. It does not take a photo by itself.