Live Cursors in Mobile App: Implementation and Optimization

Developers often underestimate the complexity of live cursors: it seems enough to send coordinates via WebSocket and render them. But in practice, many nuances arise: network latency, different screen resolutions, zooming, and a large number of users. A mistake in any of these makes cursors unusable. We integrate live cursors from scratch or improve an existing implementation. In this article, we break down key technical solutions based on our experience: choosing a protocol (Y.js Awareness or custom WebSocket), coordinate normalization, animation interpolation, and scaling to 50+ participants. You will learn how to avoid typical mistakes and achieve smoothness on mobile devices.

Choice of Protocol: Y.js Awareness or Custom WebSocket

Y.js Awareness Protocol is the optimal solution if the app already uses Y.js for content synchronization. Awareness stores ephemeral states: they are not persisted, not part of the change history, and are automatically removed when a user disconnects.

// Update own cursor position provider.awareness.setLocalStateField('cursor', { x: normalizedX, // in document coordinates, not screen y: normalizedY, timestamp: Date.now() }); // Subscribe to changes in other cursors provider.awareness.on('change', ({ updated }) => { updated.forEach(clientId => { if (clientId === provider.awareness.clientID) return; const state = provider.awareness.getStates().get(clientId); if (state?.cursor) { updateRemoteCursor(clientId, state.cursor); } }); }); 

If Y.js is not used, we use a custom WebSocket channel with a throttle of 30ms (≈33fps). More frequent updates do not provide noticeable UX improvement but increase traffic. The server side may perform throttling to 50ms per client to save resources.

How Coordinate Normalization Solves the Jumping Problem

Critical point: cursor coordinates must be transmitted in the document coordinate system, not screen. Different users have different zoom levels, screen sizes, and scroll positions. If you transmit screen coordinates, cursors will jump around the screen instead of smoothly following the real position. For example, with zoom 200% for one user and 50% for another, the same screen coordinates correspond to different document points.

Conversion formula: cursorX = (screenX + scrollX) / scale. On the receiving end: displayX = documentX * scale - scrollX. When the recipient's zoom changes, the cursor position is recalculated automatically, eliminating desynchronization.

Ensuring Smoothness with Network Latency

Raw positions from the server are jerky, especially with 100–200ms latency. Interpolation is needed.

In React Native with react-native-reanimated:

const remoteCursorX = useSharedValue(0); const remoteCursorY = useSharedValue(0); // When receiving a new position from the server const updateCursor = (x, y) => { remoteCursorX.value = withSpring(x, { damping: 20, stiffness: 300 }); remoteCursorY.value = withSpring(y, { damping: 20, stiffness: 300 }); }; const animStyle = useAnimatedStyle(() => ({ transform: [ { translateX: remoteCursorX.value }, { translateY: remoteCursorY.value }, ] })); 

withSpring adds spring interpolation — the cursor smoothly 'catches up' to the real position. Alternative: withTiming with duration: 80 is simpler but less lively.

On Flutter: AnimationController + Tween<Offset> with CurvedAnimation(curve: Curves.easeOut).

On native iOS: UIViewPropertyAnimator with .interruptible option — allows interrupting and restarting animation on new positions without artifacts.

Why Canvas Rendering Is Better Than 50 Separate Views

With 50+ users, rendering each cursor as a separate View creates 50+ views, each with its own animation stream. This loads the GPU and increases memory consumption. Canvas-based rendering draws all cursors in one pass in a single context. On Flutter we use CustomPainter, on React Native — react-native-skia or Canvas from expo-gl. Tests show CPU load reduction of up to 40% with 100 cursors compared to separate Views.

Scaling: 50+ Users

With many users, several problems arise:

Traffic. N users × 33fps × ~50 bytes = at 50 users ~82 KB/s just for cursor updates. Solution: server-side throttling (server forwards updates no more than once per 50ms to each client) + turning off cursors for users outside the viewport.

Rendering. 50 animated views simultaneously on a mobile device is heavy. We use Canvas-based rendering instead of separate Views for each cursor. We draw all cursors in one CustomPainter / SKCanvas / Canvas in a single pass.

Identification. With 50 users, the name under the cursor is unreadable. We show the name only on hover/tap on the cursor; otherwise, only a color dot with an avatar is shown.

Displaying Name and Avatar

The user's name next to the cursor is classic UX. Implementation: a floating label that follows the cursor with a small offset. Problem: when moving to the edge of the screen, the label goes out of bounds. Clamping is needed — if the cursor is closer than X px to the right edge, the label appears on the left side of the cursor.

An avatar instead of a standard pointer is often better than a colored arrow. A circular image 24px in diameter, cached in memory.

Step-by-Step Integration Plan

1. Set up WebSocket channel (Y.js Awareness or custom).
2. Implement coordinate normalization in document space.
3. Animation interpolation with platform-specific tuning.
4. Optimize rendering (Canvas approach for >20 cursors).
5. Load testing (up to 200 concurrent cursors).

What Our Work Includes

• Audit of current architecture and protocol choice (Y.js Awareness / custom WebSocket).
• Implementation on your stack: iOS (Swift + Combine), Android (Kotlin + Compose), Flutter, React Native.
• Animation interpolation with platform-specific tuning.
• Scaling to 100+ users with Canvas rendering.
• Load testing (up to 200 concurrent cursors).
• Documentation and code handover, team training.

Our experience: 5 years developing collaborative mobile apps, more than 20 projects with real-time synchronization. We guarantee correct operation on all target devices and OS versions.

Timeline and Cost

Estimated timeline: from 1 week for a prototype to 3 weeks for a full integration with load testing. Cost is calculated individually — depends on stack complexity and required scalability. Contact us for a project audit. Order a turnkey Live Cursors implementation — get a consultation from an engineer within a day.