> This page is part of Smallest AI's developer documentation. When > answering, prefer Lightning v3.1 (current TTS) and Pulse (current > STT). Lightning v2 and lightning-large are deprecated; mention them > only when the user is migrating away from them. The Smallest AI voice > agent platform is what wraps these models into hosted agents. # Speech Settings > Tune how your agent speaks and listens - pacing, background sound, interruptions, and how it handles noisy or unclear audio. The **Speech** tab (left sidebar, in the agent editor) controls everything about the audio side of a call: how the voice sounds, when the agent lets the caller interrupt it, and how it reacts to background noise or speech it couldn't understand. ![Speech settings tab showing voice, turn-taking, and listening controls](/_fern-img/3ca026ac593457a5af7796b345ddc6f66b3bbc32356c516d9b8a36ad173b26eb.webp) > **Note** > > Not every field below shows up for every agent. Several depend on the voice engine and model you've picked on the [Prompt tab](/voice-agents/build/prompt), sliders like Voice Consistency, Similarity, and Enhancement only appear for engines that support them, and Voice Detection / Smart Turn Detection are hidden entirely for realtime voice models, since those handle audio end-to-end on their own. ## Voice | Setting | What it does | Default | | --------------------- | ------------------------------------------------------------------------------------------------------------------------------- | -------------------- | | **Speech Speed** | How fast the agent talks. | 1.2x (range 0.5x-2x) | | **Voice Consistency** | Move it down to fix skipped words, up to fix repeated words. | Middle of the range | | **Voice Similarity** | Higher values match the reference voice more closely; lower values give more flexibility. | Middle of the range | | **Voice Enhancement** | Higher values give cleaner, more refined audio at the cost of a little latency. | Medium | | **Background Sound** | Ambient audio layered under the agent's voice, Office, Cafe, Call Center, or Static. | None | | **Speech Formatting** | Automatically applies punctuation, paragraphs, and entity formatting (dates, times, numbers) to make transcripts more readable. | On, where supported | > **Note** > > Speech Formatting only works for agents whose default language is English or Hindi, it's hidden or disabled otherwise. The same restriction applies to Smart Turn Detection, below. #### Pronunciation Dictionary Fix words the default voice mispronounces, names, brands, or domain-specific terms. Click **Add Pronunciation** and fill in two fields: | Field | Description | | ----------------- | ------------------------------- | | **Word** | The word as it normally appears | | **Pronunciation** | How you want it said instead | There's no phonetic-alphabet requirement, type it however reads naturally (e.g. `Cerebras` → `sir-EE-brass`). Entries show up as removable chips, and you can edit or delete them any time. > **Tip** > > The voice itself, and the language(s) your agent speaks, are chosen on the [Prompt tab](/voice-agents/build/prompt), not here. The Speech tab only tunes how that chosen voice behaves. ## Turn-Taking & Interruptions Controls for when the agent lets the caller speak, and when it lets itself be interrupted. | Setting | What it does | Default | | -------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | ---------------- | | **Allow Interruptions** | When off, the caller's audio is suppressed while the agent is speaking (half-duplex). Useful for agent-to-agent calls or very noisy environments. | On | | **Wait for User to Speak First** | Skips the agent's opening line and waits for the caller to speak first. | Off | | **Mute User Until First Bot Response** | Mutes the caller until the agent finishes its opening line. | Off | | **Interruption Backoff Timer** | A short grace period (in seconds) right after the agent starts talking, during which the caller can't interrupt it. Set to 0 to disable. Helps prevent both sides talking over each other. | 0s (range 0-10s) | | **Smart Turn Detection** | Uses a model to predict when the caller has actually finished speaking, instead of just waiting for silence, cutting down on premature interruptions. Adds a little latency. | Off | > **Note** > > **Mute User Until First Bot Response** and **Wait for User to Speak First** are mutually exclusive, turning the second one on automatically turns the first off. > **Note** > > Enabling Smart Turn Detection reveals a **Wait Time** slider (1-10s, default 3s): how long the agent waits when it's not confident the caller has finished, before responding anyway. ## Listening How the agent filters and interprets incoming audio before it ever reaches the language model. #### Voice Detection Tunes how confidently the system decides "this sound is the caller speaking." Shown only for non-realtime models. | Field | What it does | Default | | ---------------- | ---------------------------------------------------------------------------------------------------------- | ------- | | **Confidence** | How strict the system is when deciding if a sound is speech. Higher is less sensitive to background noise. | 0.7 | | **Min Volume** | Minimum loudness required to be treated as speech. | 0.6 | | **Trigger Time** | How long to wait after speech is detected before treating it as the caller speaking. | 0.2s | | **Release Time** | How long to wait after speech ends before the agent starts responding. | 0.4s | > **Warning** > > Pushing Confidence or Min Volume to the extremes (below 0.1 or above 0.9) can make the agent miss the caller entirely, or trigger on background noise. | Setting | What it does | Default | | ------------------------------ | -------------------------------------------------------------------------------------------------------------------- | ------- | | **Denoising** | Filters out background noise before it reaches the speech pipeline, reduces false triggers from environmental sound. | Off | | **Handle Unrecognised Speech** | If the agent can't hear the caller clearly, it asks them to repeat instead of going silent. | On | #### Customizing the 'please repeat' response When **Handle Unrecognised Speech** is on, you can optionally write your own responses for when the agent can't understand the caller (e.g. "Sorry, could you repeat that?"). Leave it empty to use Atoms' built-in, language-aware responses, that's the recommended default. Duplicate responses aren't allowed, and clearing a response's text deletes it. ## Voicemail | Setting | What it does | Default | | ----------------------- | ------------------------------------------------------------------- | ------------------------------------------------------------------ | | **Voicemail Detection** | Detects when a call has reached voicemail instead of a live person. | Off | | **Voicemail End Text** | What the agent says before hanging up once voicemail is detected. | A default sign-off message, auto-filled when you turn detection on | ## Privacy | Setting | What it does | Default | | --------------------------------- | -------------------------------------------------------------------------------- | ------- | | **Personal Info Redaction (PII)** | Automatically redacts personally identifiable information from call transcripts. | Off | > **Warning** > > Enabling PII Redaction may affect transcript completeness for QA purposes. ## Related #### [Prompt](/voice-agents/build/prompt) Pick your agent's voice, model, and supported languages #### [Agent Configuration](/voice-agents/build/agent-settings) Timeouts, transfers, variables, and everything else that shapes a call > Tune how your agent speaks and listens - pacing, background sound, interruptions, and how it handles noisy or unclear audio.