--- title: "Conversation Settings" nav_title: "Settings" description: "Configure title, branding, voice, access control, automations, and the main settings that control how your conversation looks and behaves." tags: ["conversation branding", "voice settings", "automated routing", "access control"] date: "2026-03-30" nav_order: 2 --- # Conversation Settings You can change most settings by asking the design agent. For example: *"Make the welcome message shorter"* or *"Use our brand colors"* or *"Require participants to sign in with Google."* You can also access settings directly by switching from the Outline view to the Settings view in the sidebar. Settings are organized into six sections. ## Title and Welcome Message The first thing participants see. #### Conversation Title The external title shown to participants. This is different from the internal name (see Advanced). #### Welcome Message The intro text participants see before the conversation starts. Explain what the conversation is about and roughly how long it takes. Up to 500 characters. You can reset to the auto-generated default at any time. ## Branding Control how your conversation's identity appears to participants. #### Interview Identity Choose between two modes: - **Personal** — Participants see your name, avatar, and an optional signature/byline (e.g., "Sarah, Product Manager at Acme"). Use this when the conversation should feel like it is coming from a real person. - **Brand** — Participants see your company logo and name. Use this for product-level conversations where the individual behind it does not matter. #### Custom Link Set a clean, branded URL slug for your conversation link (lowercase, numbers, and hyphens only). #### Custom Logo Upload separate logos for light and dark mode. When enabled, your logo is used as the conversation logo. Use **Remove Perspective Branding** to hide "Powered by Perspective" labels. #### Custom Theme Colors Set your primary color for both light and dark mode: | Color | What it controls | |-------|-----------------| | **Primary** | Buttons, progress fill, links | You can set this manually or let Perspective discover your brand color automatically from your website. ## Voice & Tone Conversations support both text and voice channels. You can enable or disable each channel independently (see Advanced). #### Voice for Conversations Select the voice used in voice-enabled conversations. Preview each voice before choosing. The selected voice affects how the agent sounds when participants use voice mode. #### Voice Instructions Describe the voice personality, accent, dialect, or speaking style for voice conversations. #### Conversation Mode Controls how voice conversations feel to participants. **Live conversation (default)** — A real-time call, like a phone call: the agent asks its questions out loud in its selected voice, and the participant answers by talking. **On-screen questions, spoken answers** — The agent's voice is muted. Its questions (and everything else it says) appear as written messages on screen, while participants still answer out loud by voice — you keep the speed and richness of spoken answers without any agent audio. Choose this when: - **You'd rather not use an AI voice at all.** If none of the available voices or accents feels right for your audience or brand, this mode simply mutes the agent — the conversation still flows, just in writing on its side. - **You want a read-at-your-own-pace experience.** Participants take the time they need with each question before answering. A few things to know: - This only affects voice conversations in the browser. Text conversations are unchanged, and phone calls are always live (a muted phone call wouldn't work). - Everything else works the same — live transcription, the conversation transcript, and answer recordings are all still captured. - With on-screen questions, the **Voice for Conversations** and **Voice Instructions** settings apply only to phone calls, since the agent never speaks in the browser. ## Language #### Conversation Language Sets the language your conversations happen in. It controls two things at once: - **What the agent says** — the agent opens and conducts the conversation in this language, in both text and voice. - **What the agent hears** — speech recognition is tuned to this language across voice conversations, phone calls, and microphone dictation in text chats. **Auto-detect (default)** — The agent adapts to each participant: it opens based on your [outline](/docs/guide/design/create-research-outlines) instructions or the participant's device language, and speech recognition detects the spoken language automatically. This is the right choice for international or mixed-language audiences. The trade-off is that speech recognition without a language hint can mishear short answers, especially in non-English languages — a brief reply is sometimes transcribed in the wrong language entirely. **A specific language** — The agent opens and conducts conversations in that language (your outline instructions can still override what it says), and speech recognition listens specifically for it, which makes voice transcription noticeably more accurate. The gains are largest for non-English languages. Choose this when you know essentially all of your participants speak one language — for example, research run in a single market. The trade-off: speech recognition is tuned to that one language, so it's less accurate for a participant who speaks a different one. **How it works alongside your outline and your participants:** - The two halves — what the agent *says* and what it *hears* — can come from different places. You can also steer the spoken language directly in your [outline](/docs/guide/design/create-research-outlines) instructions (for example, telling the agent to conduct the conversation in a particular language); those instructions take precedence over this setting for what the agent says. Speech recognition always follows the language selected here. - If a participant clearly switches to another language mid-conversation, the agent follows them in what it says — but speech recognition stays tuned to the selected language, so it still works, just less accurately for the new one. This setting decides where the conversation starts and what speech recognition leans toward, not a hard limit on what participants may speak. You can also just ask the [design agent](/docs/guide/design/create-conversation) to set this for you, or whether pinning a language makes sense for your audience. ## Automations Control what happens automatically during and after conversations. Automations are organized into two tabs. #### Per Interview Triggered after terminal conversation events. Participant-facing email automations run only after completed, non-manual conversations: - **Generate interview summary** — Always on. A summary is automatically created after each conversation. - **Generate trust assessment** — Always on. Flags suspicious patterns and gives you a trust score to review. - **Thank-you email to participants** — Optionally send an automatic thank-you email after completion. Recommended. - **Email summary when interviews complete** — Get an email summary each time a participant finishes so you can stay on top of incoming responses. - **Custom automations** — Add your own per-interview triggers for webhooks, Slack messages, CRM updates, or email sequences. See [Automations](/docs/guide/design/automations) for the design workflow and [Automation Recipes](/docs/build/automation-recipes) for implementation patterns. #### Scheduled Recurring reports on a daily or weekly cadence: - Set up digest reports that summarize new conversations over a time period. - Useful for staying informed without checking the dashboard. ## Advanced #### Internal Name A name visible only to you and your team, separate from the title participants see. Use this to organize conversations internally (e.g., "Q1 Churn Research - Enterprise"). #### Require Sign-In By default, conversations are anonymous. Toggle this on to require authentication. When enabled, you choose which sign-in providers to accept: - Google, LinkedIn, Microsoft, Email #### Voice and Text Enable or disable the two ways participants can converse: **Voice** (they talk out loud with the agent — live or with on-screen questions, per [Conversation Mode](#conversation-mode)) and **Text** (they type in a chat). At least one must be enabled. When both are on, participants can switch between them during the conversation. #### Participant Limit Set a maximum number of participants. Once reached, the conversation link stops accepting new people. Leave blank for unlimited. #### No-Activity Timeout Minutes of inactivity before the conversation is automatically reviewed and wrapped up based on the participant's progress. Default: 30 minutes. Minimum: 10 minutes. Leave blank to use the default. #### Allow Multiple Conversations When on, each participant can complete the conversation multiple times. Useful for longitudinal studies or recurring feedback. When off, each participant can only complete it once. #### Quick Reply Options When on, participants can see tappable response options for eligible closed questions such as yes/no, ratings, and pick-from-list prompts. Helps participants respond faster, especially on mobile. #### Remove Perspective Branding Hide supported "Powered by Perspective" labels from participant-facing surfaces.