Switch Android Talk Mode to realtime Gateway relay
TL;DR
Android's Talk Mode now utilizes realtime Gateway relay voice sessions, enabling streaming mic input, low-latency audio playback, and on-screen transcripts.
What changed
OpenClaw switched Android Talk Mode to a realtime Gateway relay on 19 May 2026. The update routes voice sessions through streaming mic input and low-latency audio playback.
On-screen transcripts now appear live during conversations. The change replaces earlier batch-style handling with continuous relay connections.
Why it matters
Vibe Builders gain faster voice access to their persistent self-hosted agent across WhatsApp, Telegram, and similar apps. The relay keeps memory and scheduled tasks intact while cutting response lag.
It pressures closed cloud agents on responsiveness but exposes OpenClaw users to any new relay reliability or billing variables. The bet is that local control plus quicker voice beats pure managed convenience for daily briefings and light automation.
How to use it
Pull the latest OpenClaw build through the existing CLI install on your VPS or local machine. Edit the YAML config to enable the realtime Gateway option under Talk Mode.
Connect your Android device through one of the nine supported messaging apps and start a voice session. No extra plan or payment is required beyond your current VPS and LLM token costs.
Watch for
Stable sub-300ms latency plus accurate transcripts across varied networks would validate the update. Relay outages or sudden token spikes would break the improvement. Expect a similar realtime path for iOS or desktop voice next.
Who this matters for
- Vibe Builders: Enable the realtime Gateway in your YAML config to get instant voice responses on WhatsApp.
- Developers: Integrate the new streaming mic input and low-latency relay to build responsive voice-first agents.
Harsh’s take
OpenClaw moving to a realtime Gateway relay is a necessary pivot from clunky batch processing to fluid interaction. By slashing latency and providing live transcripts, it bridges the gap between self-hosted DIY setups and polished commercial voice assistants. This update proves that open source agents can compete on speed without sacrificing the privacy of a VPS-hosted backend.
Operators should watch the token burn. Realtime streaming often leads to higher API costs compared to optimized batching. However, the trade-off for sub-300ms latency is worth it for anyone using voice as their primary interface for task management.
This move sets a high bar for other self-hosted frameworks to adopt continuous relay connections or risk feeling like legacy tech.
by Harsh Desai
About OpenClaw
View the full OpenClaw page →All OpenClaw updatesGo deeper
More AI news
- FeatureHermes Agent verifies work with completion contracts and evidence ledgers
Hermes Agent records verification evidence for coding tasks. The /goal command uses completion contracts to judge success against test runs rather than model assertions.
- FeatureCursor adds cloud agent management to the Agents window
Cursor sets up cloud development environments in under 10 minutes, spins up isolated cloud subagents using /in-cloud, and hands off sessions between local and cloud.
- FeatureCursor introduces /automate skill for automating repetitive tasks
Cursor's new /automate skill creates automations from plain language. Workflows trigger via Slack emojis or GitHub events while cloud agents access virtual computers.