> ## Documentation Index
> Fetch the complete documentation index at: https://koreai.mintlify.app/llms.txt
> Use this file to discover all available pages before exploring further.
# Voice Interactions
When integrating a voice channel through the Automation Node in AI for Service, two modes are available: **Real-Time Voice** and **TTS Streaming (ASR/TTS)**. These modes determine how user audio is processed and how responses are delivered.
```
Real-Time Voice → Multimodal LLM → Audio response
TTS Streaming → Text LLM → AI for Service TTS engine → Audio response
```
***
## Real-Time Voice
Real-Time Voice enables natural conversation using multimodal language models that process audio input and generate audio output directly.
**How it works:**
* The Platform uses the configured real-time voice model to process user audio and generate audio responses.
* Real-Time Voice must be enabled in both the AI for Service Automation Node and the Agentic App on the Platform.
* If not enabled on the Platform side, audio requests fail with errors.
* When Real-Time Voice is disabled in the Automation Node, the system defaults to ASR/TTS.
[Wait Time Experience](/agent-platform/agents/agentic-apps/settings/app-configurations#waiting-experience) is not supported for Real-Time Voice interactions.
**To configure:**
1. Enable Real-Time Voice in the AI for Service Automation Node. 
2. Enable Real-Time Voice in the Agentic App and configure a model that supports real-time voice. 
***
## TTS Streaming (ASR/TTS)
TTS Streaming is the default when Real-Time Voice is disabled. The Platform uses a text-based LLM to generate responses; AI for Service converts the text to speech using TTS engines (Deepgram, ElevenLabs).
**How it works:**
* The Platform generates a text response and sends it to AI for Service.
* AI for Service converts the text to speech using configured TTS engines.
* When TTS Streaming is **off**, the full audio response is delivered only after the complete output is generated.
* When TTS Streaming is **on**, text is streamed progressively as it is generated, reducing latency.
**Wait-time experience:** When processing is delayed, the Platform sends a filler or system-generated message as configured. AI for Service presents this to the user via TTS as it is received (when streaming is enabled).
**To configure:**
1. Disable Real-Time Voice in the AI for Service Automation Node.
2. To enable streaming, turn on the **TTS Streaming** option. 
3. No additional configuration is required on the Platform side.
***