The Behaviour tab controls your agent’s default model, session handling, and fallback responses. Settings here apply project-wide - individual playbooks can override model settings where needed.

Some settings differ depending on whether you’re building a Chat or Voice project. Differences are noted below.

Model & reasoning

Default model

The LLM that powers your agent. Select your primary model, temperature, and max token limit. These can be overridden at the playbook level.

Default model settings with temperature and token limit
Choose the default model and tune its response settings.
  • Model - your primary LLM (eg: Claude 4.6 Sonnet, Gemini 2.0 Flash)
  • Temperature - controls how creative or deterministic responses are. Lower values (closer to 0) produce more consistent output, higher values (closer to 1) introduce more variation.
  • Max tokens - maximum response length per turn

Memory

How many conversation turns the agent keeps in context (10-100, default 50). Higher values give the agent more history to work with but increase latency and token usage.

Memory is stored in the {vf_memory} variable. Most agents work well between 25-50 turns. If you’re hitting token limits or noticing slow responses, try reducing this value.

Memory turn limit slider
Set how many conversation turns the model keeps in memory.

Conversation memory that exceeds this turn count isn’t forgotten completely. Rather it’s condensed and summarized so your agent still has access to critical information, while keeping its context window small for optimal performance.

Faster processing

Enables faster processing for supported models. Uses more credits. Off by default. When on, you can check any model dropdown in Voiceflow to see supported models.

Priority processing toggle
Enable priority processing for supported models when lower latency matters.

Outage protection

Voiceflow automatically detects model provider outages and allows you to switch to a fallback LLM provider if your primary provider goes down. Configure a fallback for each provider independently - for example, route OpenAI failures to Claude, and Anthropic failures to GPT-4.1.

Outage protection settings mapping providers to fallback models
Choose fallback models the agent can use during a provider outage.

Note: your agent can perform very differently when being powered by different models.

Timezone

Sets the timezone for the {vf_now} and other built-in time variables, and any time-aware behavior.

Timezone dropdown in Behaviour
Set the timezone used by the agent’s time-based behavior.

You can think of this as your agents internal clock. You also have access to the built-in variable{vf_user_timezone}.

Voice output

Voice output is also available in chat projects as Voiceflow’s chat widget handles multi-modal usecases with voice mode.

Provider, voice, and model

Click the voice under Voice, or the voice at the top of the agent editor, to open the voices editor. Its library lists every provider’s voices (ElevenLabs, Cartesia, Rime, Soniox, Google, Amazon, Microsoft): search by name or voice ID, filter by language, accent, provider, model, and gender, play a sample, and click Use to pick one. The panel beside the library sets the voice’s synthesis model and its settings. You can also connect a custom voiceID from ElevenLabs if you have one.

This is your agent’s primary voice: it speaks every language that has no language override.

Language overrides

Language overrides give your agent a native voice for each language it speaks. As the agent writes each message, it picks the voice from the conversation’s {locale} variable, so a caller from France can hear your French voice while everyone else hears your primary voice.

To add one, click Add language override and choose a language (French, fr) or a regional variant (French (Canada), fr-CA). The override starts with a voice that speaks the language natively, and the voices editor opens on it so you can change the voice, its model, and its settings. Each override keeps its own settings. To remove one, hover it in the voices editor and choose Remove override from its menu.

How the voice is picked. The agent looks for an override for the exact locale first, then for the locale’s language, and otherwise uses the primary voice. Case and _ are ignored, so fr_ca matches fr-CA.

An override for a language (fr) covers every region that speaks it. Add one for each language, then add regional overrides (fr-CA) only where you want a particular accent.

Where {locale} comes from:

  • Chat widget - the user’s browser language.
  • Phone calls - inferred from the caller’s number: the main language of its country (en-US, es-MX, fr-FR), or fr-CA for a Quebec area code. A withheld number falls back to your agent’s number. If neither number identifies a country, {locale} is empty.
  • Your agent - set {locale} yourself, for example with a Set step when a user asks to switch languages. Messages written after the change use the new locale’s voice; messages already written keep theirs.

Overrides change only the voice. Speech recognition keeps the language set under Voice input, and the pronunciation dictionary applies to every voice. Agents running on a speech-to-speech model speak with that model’s own voice, so overrides don’t apply to them.

A warning icon appears next to an override whose voice doesn’t speak its language. Soniox voices speak many languages: pick which one under Language in the voices editor. Picking a Soniox voice for an override sets it to the override’s language.

Language overrides

Choose a different speaking voice for each language or regional locale in the Voices editor. The conversation’s {locale} variable selects an exact locale override first, then a language-wide override, then the primary voice.

In Voice output, click Add language override, select the language and choose its voice and model. The widget uses the browser’s language preference; phone calls infer an initial locale from the caller’s number. API integrations supply the conversation variable themselves.

Follow Set voices by language for the complete setup, regional matching rules and tests. Voice selection controls speech synthesis; configure response language and speech recognition separately.

Sync audio and text output

Streams TTS in real time, keeping audio synced with the text output. Off by default.

Voice output settings with Sync audio and text output
Choose a voice and control whether text appears in sync with speech.

Voice tuning

Fine-tune how the voice sounds:

  • Similarity boost (0-1, default 0.75) - how closely the output adheres to the source voice. Higher values are more accurate but may reduce naturalness.
  • Talking speed (0.7-1.3, default 1.00) - playback speed of the voice.
  • Stability (0-1, default 0.50) - higher values produce more consistent output across regenerations but can sound more monotone. Voice settings for speed, stability, and similarity

Pronunciation dictionary

The pronunciation dictionary rewrites words or phrases in your agent’s responses before they’re sent to text-to-speech. This ensures correct pronunciation of names, places, or technical terms:

  • enforce brand pronunciations (nginx → engine x)
  • expand abbreviations (Dr. → Doctor)
  • spell out acronyms (TTS → Text to Speech)
  • normalize locale spellings (colour → color)

Matching is case-insensitive and respects whole-word boundaries. The JSON view is used for bulk editing of large dictionaries.

For finer control you can use TTS provider-specific phonetic markup in the to field.

  • Cartesia uses a proprietary IPA syntax <<phoneme|phoneme|...>> (guide)
  • ElevenLabs (Turbo v2 / Multilingual v2 only) uses SSML <phoneme> tags with either IPA or CMU Arpabet (guide)
Pronunciation dictionary setting with an Edit button
Edit pronunciation rules for words your agent needs to say correctly.

Background audio and audio cues

Audio cues is also available in chat projects as Voiceflow’s chat widget handles multi-modal usecases with voice mode.

Add optional ambient audio or audio effects during conversations.

  1. Background audio - ambient audio that plays in the background for the duration of a call
  2. Audio cue - a subtle sound effect that plays when the agent starts thinking Background audio and audio cue selectors

Voice input

Voice input is also available in chat projects as Voiceflow’s chat widget handles multi-modal usecases with voice mode.

Provider, model, and language

Select your STT provider (eg: Deepgram, Gladia, Cartesia, AssemblyAI, Google), transcription model (eg: Flux), and language.

We highly recommend using the Deepgram Flux model. It’s incredibly strong at understanding when the user is done talking, which can massively reduced the perceived latency of voice conversations.

Keywords

Comma-separated list of words to boost recognition for - proper nouns, product names, or industry-specific terms that standard transcription might miss.

Good keyword examples:

  • Product and company names: Brand names, service names, competitor names
  • Industry-specific terminology: Medical terms (tretinoin, diagnosis)
  • Multi-word phrases: Common phrases in your domain (account number, customer service)
  • Proper nouns: Names, brands, titles with appropriate capitalization (Deepgram, iPhone, Dr. Smith) Voice input settings with provider, model, language, and keywords

End-of-turn detection

Controls when the agent decides the user has finished speaking:

  • End-of-turn confidence (0.5-0.9, default 0.60) - minimum confidence score required to close a turn. Lower values make the agent more responsive, higher values wait for more certainty.
  • End-of-turn timeout (0.5-5s, default 0.60s) - how long to wait after the user stops speaking before closing the turn, regardless of confidence. Increase this if users are getting cut off mid-sentence. End-of-turn confidence and timeout settings

Keypad input

Allow users to input digits via keypad (DTMF) during a call. When enabled, configure:

Keypad input is only available in voice projects.

  • Timeout (0-10s, default 2) - how long to wait before processing the input. Set to 0 to process only after a delimiter is pressed.
  • Delimiter - the key that signals input is complete. Options: Pound (#) or Star (*). If both a delimiter and timeout are set, whichever comes first triggers processing. Keypad input settings with timeout and delimiter controls

Initialization

Initialization workflow

A workflow that runs before the user’s first turn - to load data, authenticate, or send a static greeting before your agent takes over. It’s optional and can be left empty. Pick an existing workflow from the dropdown, or click Create workflow to build a new one. See Initialization workflow for setup and common patterns.

Initialization workflow selector in Behaviour
Choose the workflow that runs before the conversation starts.

Session & timeout

Session timeout (Chat projects)

Time in minutes before an inactive chat session ends (1-2,880 min (2 days), default 15 minutes). Toggle off to keep sessions alive indefinitely.

Chat session timeout toggle and inactivity duration
Set how long an inactive chat session remains open.

Session timeout (Voice projects)

How long to wait before a call is automatically ended due to inactivity (10-300 seconds, default 60).

Voice session timeout slider
Choose when an inactive phone conversation should end.

Chat persistence

Controls how and whether chat history is stored between sessions.

Chat persistence is only available in chat projects.

  • Never forget - conversation history persists indefinitely. The user can return at any time and pick up where they left off.
  • Forget after all tabs are closed - history is cleared when the user closes all browser tabs with the chat widget. Reopening starts a fresh session.
  • Forget after page refresh - history is cleared every time the page reloads. Each page visit is a new conversation. Chat persistence setting with Never forget selected

Events

Events let you trigger agent behavior from outside the conversation - a webhook, a CRM update, a system alert, or any external signal your application sends.

Event mappings with trigger and action fields
Map incoming events to the action the agent should take.

To create an event, click New event, give it a name, and choose its behaviour:

Fallback

Global no match and Global no reply settings
Configure fallback messages for unmatched requests and silent users.

Global no-match

How the agent responds when it can’t match any skill, user intention, or knowledge base result. Click Edit to configure:

  • Generative - provide a prompt that tells the agent how to handle the situation. The agent generates a response dynamically based on your instructions.
  • Scripted - write a fixed response that’s sent every time. Use this when you want exact control over the wording.

    If you’re building a fully agentic project, you don’t need to worry about this behaviour setting. It’s really for experiences that don’t rely on an agent to handle un-happy path dialogs.

Global no-reply

How the agent responds when the user says nothing. Off by default. When enabled, click Edit to configure:

  • Generative - provide a prompt for how to re-engage the user (eg: “Ask them if they’re still there, or nudge them toward the conversation’s goal”).
  • Scripted - write a fixed follow-up message.
  • Inactivity time - how many seconds of silence before the agent responds (default 10).
  • Max no-reply messages per turn - how many times the agent will follow up before ending the session (default 5). For example, with 10 seconds and 5 attempts, the session ends after 50 seconds of total silence.

Use the up and down arrow keys to select a result, Enter to open it, and Escape to close the search.