Voicebot interactions can experience delays when the Text-to-Speech (TTS) engine generates the complete audio response before playback begins. To reduce this delay, Voicebots now support streaming TTS with ElevenLabs, allowing audio playback to start as soon as the first audio segment is available.

Streaming TTS improves responsiveness, reduces perceived wait times, and creates more natural voice interactions.

Why Use Streaming TTS?

In traditional TTS workflows, the system generates an entire audio file before sending it to the Voicebot. For longer responses, this can introduce noticeable delays before customers hear the bot speak.

With streaming enabled, audio is generated and delivered in smaller chunks, enabling playback to begin almost immediately. This results in:

• Faster response times
• Reduced perceived latency
• More natural conversational flow
• Improved customer experience

Enablement

This capability may require additional configuration. Contact the product team for enablement assistance.

Behaviour Without Streaming

When streaming is not enabled:

1. The TTS engine receives the response text.
2. The complete audio file is generated.
3. The Voicebot waits for audio generation to finish.
4. Playback starts only after the full audio file is available.

This can increase response latency, particularly for longer responses.

Behaviour With Streaming Enabled

When streaming is enabled:

1. ElevenLabs generates audio in real time.
2. Audio chunks are sent to the Voicebot as they become available.
3. Playback begins immediately after the first audio segment is received.
4. Audio continues to stream until generation is complete or interrupted.

This significantly reduces the delay between text generation and speech playback.

Key Capabilities

Real-Time Audio Streaming

Audio is streamed in smaller segments, allowing playback to begin before the full response is generated.

ElevenLabs Streaming Integration

Uses the ElevenLabs streaming API to support low-latency speech generation.

Continuous Voicebot Playback

Audio chunks are delivered and played seamlessly throughout the interaction.

Graceful Stream Handling

The streaming session ends cleanly when speech generation completes or when the interaction is interrupted.

Automatic Fallback

If streaming becomes unavailable because of network issues, API errors, or other failures, the system automatically falls back to standard TTS processing.

Monitoring and Logging

Streaming activity is logged to assist with troubleshooting and performance monitoring, including:

• Stream start and end events
• Audio chunk delivery latency
• Stream interruptions and failures
• Playback readiness events