Live Captions
Live Captions uses AI to generate real-time captions for videos, audio, and calls on Windows Copilot+ devices with on-device NPU processing.
Why It Exists
Users need real-time text captions for video and audio content, especially for accessibility, noisy environments, or silent viewing.
How It Works
Live Captions is Microsoft's AI-powered real-time captioning system for Windows Copilot+ devices. Using on-device NPU processing and speech recognition models, the system generates captions for videos, audio playback, video calls, and live speech. Captions appear as an overlay that can be customized for size, color, position, and language. The system works across all apps and websites, providing accessibility and convenience without requiring internet for basic captioning.
Everyday Use Cases
- Video captioning
- Audio captioning
- Call transcription
- Live speech captions
- Accessibility support
- Silent viewing
- Multilingual captions
User Workflow
- User enables Live Captions in Windows settings
- System detects audio output or microphone input
- AI transcribes speech in real-time using NPU
- Captions are overlaid on screen
- User can customize caption appearance
AI Processing Flow
Audio or system audio is captured and processed by on-device NPU speech recognition models. The system identifies speech, separates languages, and generates captions. Multiple languages are supported through on-device models. For unsupported languages or complex audio, cloud-based models may be used with consent. Caption overlay is rendered by Windows.
Inputs / Outputs
Inputs:
- Audio input stream
- System audio
- Speech content
- Language identification
- Caption position
- User preferences
Outputs:
- Real-time captions
- Multi-language support
- Customizable overlay
- Searchable captions
- Exportable transcript
- Accessibility controls
Known Limitations
- Accuracy varies with audio quality and accents
- Limited languages for on-device processing
- Background music may interfere
- Video file processing requires additional time
Unsupported Scenarios
- Offline captioning
- Custom caption themes
- Caption import/export
- Recording transcription
- Integration with hearing aids
Performance Notes
Speech recognition in real-time with under 200ms latency; Supports 20+ languages on-device; Caption rendering within 100ms; Battery impact under 4 percent per hour; NPU processing under 3 percent
Available On
Shows where this feature is available and how its AI processing works on each platform. Availability may vary by device.
| Platform | Execution | Offline | Cloud | OS / Software |
|---|---|---|---|---|
| Copilot+ | Local | Yes | No - on-device | Windows 11 24H2 |
Research Status
Confidence and verification reflect how complete documentation is. Fields may show Not assessed or Not yet verified while research is ongoing - this flags gaps, not product deficiencies.
| Research Status | Verified |
| Confidence | High |
| First introduced | 2024-09-04 |
| Last updated | 2026-08-16 |
| Last verified | 2026-08-16 |
Research Notes
Available on Copilot+ PCs with Windows 11; Requires NPU with 40+ TOPS; Introduced with Windows 11 2024 Update; Microsoft Translator integration
Details
| Vendor | Microsoft |
| AI Platform | Copilot+ |
| Category | Translation |
Capability Mapping
This feature maps to canonical capability family: Translation.
Canonical capability: Spoken Language Translation.
- Spoken Language Translation - Translates live spoken conversation between languages in real-time.
Grounded in 8 device(s) on Copilot+.
Related Features
- Live Translation - Cross-platform equivalent
- Live Translate - Cross-platform equivalent
This page describes platform-level capability; not device-level support for any specific product.