AI Enhancement Features Explained
Learn how to use AI to improve your stream quality automatically — video, audio, product detection, and real-time optimizations.
Our AI features run in low-latency mode to improve viewer experience without blocking your stream. They are designed to be safe-by-default and respect privacy and on-device processing where supported.
Core AI Features
Auto Video Enhancement
Automatically adjusts brightness, contrast, sharpness and color balance to keep your stream looking professional across lighting conditions.
- Real-time exposure and color correction
- Adaptive sharpening for low-detail cameras
- Dynamic tone mapping to preserve highlights and shadows
Noise Reduction
AI filters reduce background noise and hum while preserving voice clarity — works on mobile and desktop microphones.
- Adaptive noise gating
- Wind & hum suppression
- Low-latency processing
Smart Cropping & Auto Framing
Keeps the subject and products in-frame by analyzing motion and repositioning/cropping intelligently so viewers always see the important parts.
- Face & product tracking
- Safe-area aware cropping for mobile
- Smooth transitions between frames
Product Recognition & Tagging
Detects and tags visible products during a stream to surface buy buttons, product cards, and searchable timestamps automatically.
- Automatic SKU linking
- Timestamped product recognition
- Overlay product cards
Auto Subtitles & Transcription
Live automatic speech-to-text provides captions in real time (multi-language support), improving accessibility and retention.
- Low-latency ASR
- Multi-language captions
- Editable post-stream transcripts
Real-time Recommendations
Use viewer engagement signals to surface product recommendations, offers, or prompts during the stream to boost conversions.
- Engagement-driven suggestions
- A/B testable prompts
- Conversion tracking hooks
We prioritize user privacy and real-time performance:
- On-device processing where supported to limit cloud uploads.
- Configurable feature levels (Balanced / Performance / Quality) to trade off CPU/network usage.
- Audit logs and opt-out controls for personally identifiable processing (face recognition, etc.).
AI features can be enabled from the Stream Settings page or via our API/SDK for programmatic control.
// example (pseudo) SDK usage
runash.stream.enableAI({
autoVideo: true,
noiseReduction: 'balanced',
smartCrop: 'beta',
})- Test features in a private stream before enabling for public broadcasts.
- Use the Performance profile on mobile devices to avoid overheating or dropped frames.
- Enable product-recognition only when you have linked SKUs and images to reduce false positives.
- Monitor CPU and network usage; adjust AI quality when necessary.
Track improvements using A/B testing and these metrics:
- Viewer retention and average view time
- Conversion rate during promoted segments
- Engagement (chat messages, reactions)
- Post-stream watch time and clip saves
Lower AI quality or switch to Performance profile; offload processing to cloud if available.
Try a different enhancement profile, or disable sharpening for noisy cameras.
Improve SKU images and metadata, and enable stricter confidence thresholds in settings.
Was this page helpful?
Your feedback helps us improve RunAsh docs.