### Answer There are a few settings you can try to improve fluency at the beginning of calls: ## Begin Message Delay You can use

begin_message_delay_ms

to delay the first message, giving the system more time to prepare before speaking. [(1) ]() [(2) ]() This setting delays the first message by the specified amount of milliseconds (valid range is 0–5000ms) , so it gives the user more time to prepare to take the call. [(1) ]() ## Voice Stability Settings - **`voice_temperature`**: Controls how stable the voice is, with values ranging from 0 to 2. A lower value means more stable speech, while a higher value means more variant speech generation. This currently applies to 11labs voices. [(1) ]() [(2) ]() - **`voice_speed`**: Controls the speed of voice (range 0.5–2) . If the agent is speaking too fast initially, lowering this may help. ## TTS and Model Optimization For latency and initial speech quality issues, consider: - Using a faster TTS (text-to-speech) system for quicker, smoother audio output. [(3) ]() - Using a faster LLM to reduce response time. ## Voice Model Selection Try different voice models via the

voice_model

setting to see if another model performs better for Hungarian. [(1) ]() [(2) ]() --- I don't have specific documentation about Hungarian language performance or known issues with initial sentence fluency. If these adjustments don'

t resolve the issue, I'd recommend reaching out to Retell support for further assistance. Check out these potentially useful sources: [Update Voice Agent]() , [Speech Processing]()

​ To learn more, see the sources I considered: [[1] Get Agent Versions](https://docs.retellai.com/api-references/get-agent-versions) [[2] Update Voice Agent](https://docs.retellai.com/api-references/update-agent) [[3] Performance-related configurations and noticeable latency during live calls](https://community.retellai.com/t/performance-related-configurations-and-noticeable-latency-during-live-calls/747) ​

Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.

@AI Evy It's gotten a little better, but he says numbers, like 24990, very poorly. It's like he's eating something very sticky.

Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.

### Answer For numbers being spoken poorly (sounding "sticky" or unclear) , try enabling

normalize_for_speech

. [(1) ]() When set to

true

, this will normalize numbers, currency, dates, etc. to their spoken form for more consistent speech synthesis. For example, it will convert "24990" to "twenty four thousand nine hundred ninety" before audio generation, which often produces clearer speech than having the TTS interpret raw numbers. You can also try: - Lowering

voice_temperature

to a value like 0.5 or 0.7 for more stable, consistent speech output. , [(2) ]() - Adding a pronunciation dictionary to guide how specific numbers or terms should be pronounced. [(1) ]() , [(2) ]()

json
"pronunciation_dictionary": [
  {
    "word": "specific_word",
    "alphabet": "ipa",
    "phoneme": "pronunciation_guide"
  }
]

- Adjusting

voice_speed

(range 0.5–2) — slowing it down slightly may help with clarity on numbers. [(1) ]() If these adjustments don't resolve the issue for Hungarian, you may want to experiment with different voice models to find one that handles numbers better in your language.

​ To learn more, see the sources I considered: [[1] Update Voice Agent](https://docs.retellai.com/api-references/update-agent) [[2] Voice inconsistency - stuttering and filler sounds (aaaa, ehhh, hmmm)?](https://community.retellai.com/t/voice-inconsistency-stuttering-and-filler-sounds-aaaa-ehhh-hmmm/769) ​