| Device | Supported Visual & Audio Context | High-Value Use Cases |
|---|---|---|
| iPhone & iPad | Camera snap, Photo Library image, or iOS screenshot | Foreign signage/menus, whiteboard photos, mobile UI errors, printed handouts |
| Mac | Chosen screen or window capture + system audio + file attachments | Screen-shared slides, live terminal/IDE debugging, spreadsheet review |
Combining Sight and Voice in One Session
During any active cloud session in GPT Realtime Assistant, attaching a screenshot or camera photo places that visual context directly into the conversation state.
Instead of typing a five-line prompt explaining layout and labels, you just speak naturally: 'Compare the two pricing tiers in this table' or 'What does the handwritten note in the margin say?'
Desktop Window Capture + System Audio on Mac
On macOS, GPT Realtime Assistant goes a step further: it can capture a specific window or display, optionally record the session locally, and listen to Mac system audio alongside your microphone.
When reviewing a recorded webinar, design critique, or technical presentation where screen sharing is permitted, the assistant sees the slide deck and hears the discussion simultaneously—saving screenshots or clips as evidence attached to follow-up tasks.
Privacy Discipline Before Capturing Screens
Any image or screen frame you attach to an online session is transmitted from your device to OpenAI's API so the model can inspect it. Close password managers, banking tabs, private patient/client identifiers, and confidential workspaces before capturing a screen, and always follow your workplace and jurisdiction's rules on recording or AI assistance.
Pre-Session Checklist
Saved locally in your browser so you can check your setup before starting.
Frequently Asked Questions
Does Arte Soft store the screenshots I attach to a session?
No. Attachments stay locally on your device with your session and are sent directly from your device to OpenAI's API during the session; Arte Soft's backend never stores your attachments, audio, or transcripts.
Can I attach screenshots as evidence on tasks after the session?
Yes. Tasks in GPT Realtime Assistant support screenshot and video evidence stored locally on your device so you remember the exact visual state behind an action item.
Continue Exploring
How Realtime Voice AI Credits Work: What Audio, Text, and Translation Actually Spend
Transparent breakdown of GPT Realtime Assistant's 2,000 monthly AI credits, why live audio costs more than text, how pre-session reservations refund, and how to stretch your balance.
Read session guide →
Private Offline AI Sessions on Apple Silicon Mac with Local Whisper and Qwen
Run completely offline transcription and AI answers on an Apple Silicon Mac using Whisper Large v3 Turbo and Qwen 3 8B—zero cloud audio, zero AI credits spent.
Read session guide →