> ## Documentation Index
> Fetch the complete documentation index at: https://docs.itellico.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# AI Pipeline Guide

> Understand how transcriber, AI model, voice, timing, and pronunciation work together as one pipeline

Your callers do not experience your transcriber, voice, model, and timing settings separately.

They experience one conversation.

This guide helps you treat the AI pipeline as one system so you can make better tradeoffs before launch.

## Why This Matters

Most voice-quality problems are not caused by one bad setting.

They usually come from the interaction between:

* how accurately the caller is transcribed
* how quickly the model decides what to say
* how natural the chosen voice sounds
* how the system handles pauses, interruptions, and pronunciation

If you only optimize one layer, the conversation can still feel slow, robotic, or error-prone.

## The Five Parts Of The AI Pipeline

| Part               | What it controls                                     | Main doc                                                                                                                    |
| ------------------ | ---------------------------------------------------- | --------------------------------------------------------------------------------------------------------------------------- |
| **Transcriber**    | How caller audio becomes text                        | [Transcriber](/build/voice-speech/transcriber)                                                                              |
| **AI model**       | How the agent reasons and responds                   | [Choose AI Model](/build/voice-speech/choose-ai-model)                                                                      |
| **Voice**          | How the response sounds to the caller                | [Select Voice](/build/voice-speech/select-voice)                                                                            |
| **Voice behavior** | Speed, stability, style, and pronunciation           | [Voice Settings](/build/voice-speech/voice-settings) and [Custom Pronunciations](/build/voice-speech/custom-pronunciations) |
| **Timing**         | Interruptions, pauses, silence, and turn-taking feel | [Turn-Taking and Timing](/build/advanced/turn-taking-and-timing)                                                            |

## Start With The Outcome You Need

Choose the pipeline configuration based on the actual conversation you are deploying.

<AccordionGroup>
  <Accordion title="Fast phone support or triage" defaultOpen>
    Prioritize low latency, clear pronunciation, and interruption handling.

    Start with:

    * a fast [transcriber](/build/voice-speech/transcriber)
    * a clear, neutral voice
    * conservative [turn-taking](/build/advanced/turn-taking-and-timing) tuning
    * minimal ambient effects
  </Accordion>

  <Accordion title="Brand-sensitive concierge or sales experience">
    Prioritize warmth, brand fit, and consistent pacing.

    Start with:

    * a voice that matches your tone and audience
    * stronger [voice prompting](/build/conversation/prompt-engineering-guide)
    * pronunciation rules for product and company names
    * test calls with realistic objections and interruptions
  </Accordion>

  <Accordion title="Multilingual or regional deployment">
    Prioritize language coverage and locale accuracy.

    Start with:

    * language support in the [transcriber](/build/voice-speech/transcriber)
    * locale-matched voices in [Select Voice](/build/voice-speech/select-voice)
    * test scripts for each target language
    * explicit prompt instructions if tone or phrasing changes by region
  </Accordion>

  <Accordion title="Compliance-sensitive or privacy-sensitive workflows">
    Prioritize clarity, consent, and predictable behavior.

    Start with:

    * short, direct voices with minimal embellishment
    * clear [announcements](/build/advanced/announcements)
    * explicit [privacy controls](/build/advanced/conversation-privacy-controls)
    * conservative timing settings so callers can interrupt easily
  </Accordion>
</AccordionGroup>

## Configuration Order

Work through the pipeline in this order. Each layer depends on the one before it.

| Step                     | What to configure                                        | Why first                                                 |
| ------------------------ | -------------------------------------------------------- | --------------------------------------------------------- |
| **1. Transcriber**       | Language, provider, model                                | If the caller is misheard, nothing downstream can recover |
| **2. Voice**             | Provider, voice, cloning                                 | Pick what callers hear once transcription is solid        |
| **3. Voice refinements** | Settings, pronunciations, ambient sound, thinking sounds | Fine-tune after the core voice is chosen                  |
| **4. Timing**            | Turn-taking, silence, interruptions                      | Tune last — timing sliders can mask deeper problems       |

<Tip>
  Do not start with timing. If the transcriber, voice, or prompt is already causing friction, adjusting timing hides the real problem instead of fixing it.
</Tip>

**Where to go for each step:**

* [Transcriber](/build/voice-speech/transcriber) — provider and language breakdown
* [Select Voice](/build/voice-speech/select-voice) — catalog and provider guidance
* [Voice Settings](/build/voice-speech/voice-settings), [Custom Pronunciations](/build/voice-speech/custom-pronunciations), [Ambient Sound](/build/voice-speech/ambient-sound), [Thinking Sounds](/build/voice-speech/thinking-sounds)
* [Turn-Taking and Timing](/build/advanced/turn-taking-and-timing)

## Common Symptoms And Where To Look First

| Symptom                                         | First place to look                                                | Then check                                                               |
| ----------------------------------------------- | ------------------------------------------------------------------ | ------------------------------------------------------------------------ |
| Agent mishears names, addresses, or numbers     | [Transcriber](/build/voice-speech/transcriber)                     | [Custom Pronunciations](/build/voice-speech/custom-pronunciations)       |
| Voice sounds wrong for the brand                | [Select Voice](/build/voice-speech/select-voice)                   | [Prompt Engineering Guide](/build/conversation/prompt-engineering-guide) |
| Speech sounds robotic or uneven                 | [Voice Settings](/build/voice-speech/voice-settings)               | [Select Voice](/build/voice-speech/select-voice)                         |
| Agent cuts callers off                          | [Turn-Taking and Timing](/build/advanced/turn-taking-and-timing)   | [Transcriber](/build/voice-speech/transcriber)                           |
| Agent feels slow after the caller stops talking | [Turn-Taking and Timing](/build/advanced/turn-taking-and-timing)   | [Choose AI Model](/build/voice-speech/choose-ai-model)                   |
| Product names or company names are spoken badly | [Custom Pronunciations](/build/voice-speech/custom-pronunciations) | [Prompt Engineering Guide](/build/conversation/prompt-engineering-guide) |
| Cloned voice sounds inconsistent                | [Voice Cloning](/build/voice-speech/voice-cloning)                 | [Voice Settings](/build/voice-speech/voice-settings)                     |

## A Practical Rollout Sequence

<Steps>
  <Step title="Prove the logic in chat">
    Confirm the prompt, tools, and knowledge work before you spend time on voice tuning.
  </Step>

  <Step title="Evaluate the voice in Web Call">
    Listen for pace, pronunciation, and interruption feel in the browser.
  </Step>

  <Step title="Validate the full call on the phone">
    Run at least one real phone call. Phone audio and network behavior often change the result.
  </Step>

  <Step title="Review the conversation detail">
    Check the transcript, timing, tool execution, and any post-call automation before launch.
  </Step>
</Steps>

## Common Mistakes

<AccordionGroup>
  <Accordion title="Choosing the prettiest voice before checking transcription" icon="microphone">
    A beautiful voice does not help if the caller is transcribed inaccurately. Start with recognition quality, then optimize style.
  </Accordion>

  <Accordion title="Trying to fix slow responses with ambient sound" icon="music">
    Ambient sound can improve feel, but it does not solve slow model responses, slow tools, or high-latency transcription.
  </Accordion>

  <Accordion title="Testing only with your own voice and accent" icon="users">
    Always test with the kinds of callers you actually expect: different accents, speeds, noise levels, and interruption patterns.
  </Accordion>

  <Accordion title="Changing multiple layers at once" icon="sliders">
    If you change the transcriber, voice, prompt, and timing together, you will not know what actually improved or broke the conversation.
  </Accordion>
</AccordionGroup>

## Next Steps

<CardGroup cols={2}>
  <Card title="Select Voice" icon="microphone" href="/build/voice-speech/select-voice">
    Browse, preview, and choose the voice your callers hear
  </Card>

  <Card title="Transcriber" icon="ear" href="/build/voice-speech/transcriber">
    Pick the speech-to-text layer that fits your languages and latency needs
  </Card>

  <Card title="Voice Cloning" icon="wand-magic-sparkles" href="/build/voice-speech/voice-cloning">
    Create and evaluate custom branded voices
  </Card>

  <Card title="Turn-Taking and Timing" icon="wave-square" href="/build/advanced/turn-taking-and-timing">
    Tune pauses, interruptions, and silence handling
  </Card>
</CardGroup>