Voice Screen and On-Screen Content

Open in CMS

On this page you decide what users see while they talk to your agent: a full-screen immersive view or a transcript, attachments during a call, which content appears on screen (the canvas) instead of being spoken, subtitles in the immersive view, where the immersive view shows what the assistant is working on, and how a new immersive call opens.

All of these settings are on the Agent voice page (Conversation β†’ Voice in the CMS sidebar), which is one page of sections, and appear only when Voice conversations is on. The chips at the top of the page jump to each section.

In this example Atlas, Kirana Group's operations assistant, opens calls in the Immersive view, and tells the assistant to put approval chains and project status on screen as cards while it speaks only the headline.


Before you start


Voice screen style

The Screen style field at the top of the Call screen section. It controls how a voice session is presented the moment it starts. Everything in the fold Immersive view options under it (opening sequence, subtitles, progress indicator, immersive view instructions) applies to the immersive view only; when you pick Transcript, one note under the style says they only reach users who switch to the immersive view themselves.

Call screen section for Atlas: Screen style Immersive (the view a call opens in; users can change it anytime), and the closed fold Immersive view options for opening sequence, subtitles, progress, instructions

Atlas keeps Immersive, the default.

ValueWhat the user sees
Immersive (default)A full-screen, voice-only view: a sound-wave animation reacting to who is speaking, no visible transcript text. Charts, tables, and cited sources produced during the call appear as swipeable pages in a canvas pager instead.
TranscriptA text-style view: the spoken conversation is shown as a scrolling transcript of chat bubbles, same as a regular chat thread, with clickable citations and a text box to type into mid-call.

This is only the starting point: the end user can still switch to the other display from their own preferences once a call is underway.

When an immersive call opens, the caller sees the agent's logo in the middle of an otherwise empty screen; no welcome text is shown under it. The header shows the agent's name and its state (Listening while it waits for the user). At the bottom are the end call button, the microphone button, and the attach button, with arrows on the right to page through what appears on the canvas later.

Immersive voice call with Atlas: the header shows the Atlas logo and name with the status Listening and a settings button, the center shows the Atlas Kirana Group Operations logo in a circle, and the bottom bar shows Powered by Qlar, the end call, microphone and attach buttons, and previous and next page arrows

Start screen

Found in the Start screen section, in the Opens on field, while Voice conversations is on; the overview at the top of the page shows the current choice. Decides what your agent's own link (app.qlar.ai/<agent-code>) opens:

ValueWhat the visitor gets
Text chat (default)The chat, as always. A voice call starts from the + menu.
VoiceStraight into a voice call, shown in the voice screen style above. The text chat never shows on the way in. Ending the call, or Prefer typing? Open chat on the waiting screen, leads to the text chat.

The setting only changes the bare link opened in a browser tab or the installed app. These always open text chat: website widgets (unless they ask for voice with data-start-screen="voice"), the preview in the CMS, links from WhatsApp Redirect to web, and notification links to a conversation.

Under the choice, two links with a copy button open one screen whatever the start screen is:

LinkOpens
app.qlar.ai/<agent-code>/chatThe text chat.
app.qlar.ai/<agent-code>/voiceA voice call. If voice conversations are off for the agent, it opens the text chat instead.

Both links show your agent's name and picture when shared in WhatsApp, Slack or social media, like the bare link.

Tap to start

A call opened from a link starts by itself when the browser allows sound, for example in the installed app or for a visitor who uses your agent often. Otherwise the voice screen shows Tap to start on the logo: browsers, iPhone in particular, play no sound before the visitor has touched the page. Nothing starts before that tap, so the microphone is not requested and no call time is used for a visitor who only looks.


Attachments during an immersive call

In the immersive view, an attach button sits right of the microphone. Users can pick an image or a document (on a phone: Camera, Photo, or Files), and it appears as a small preview above the call controls β€” the same boxes the chat input shows.

There is no send button in a voice call, so an attachment waits: it goes to the assistant together with the next thing the user says (or, in push-to-talk, when they release the talk button), and the preview clears once it has been sent. Until then the user can remove it. The assistant then answers that spoken question with the file in view. For example, a site supervisor asks Atlas "what's the PO number on this delivery note?" with a photo of the delivery note attached.

What can be attached is controlled on Model & Context β†’ Attachments: the same Allow image upload, Allow document upload, and Document upload restrictions that apply to chat, checked the same way when the file is uploaded.

A few differences from chat worth knowing:

  • Images depend on the conversation style. With Turn-by-turn they always work: the voice reads images itself, whichever conversation model the agent uses for chat. In Natural conversation the voice can't see images β€” an image goes to the Answering AI instead, and only some Answering AI models take images in a call. When they can't, the attach button offers documents only.
  • Documents are read as text, exactly as in chat. Because a voice session holds far less context than a text conversation, only roughly the first 12,000 characters of a document (24,000 across the files sent together) reach the assistant. It is told when a document was shortened, so it can say the answer may be in the part it didn't see. The full text is still recorded in the conversation. In Natural conversation the text goes to the Answering AI: roughly the first 8,000 characters of each document for Qlar's built-in models, while custom models receive documents the way they do in chat.
  • Attachments are recorded on the conversation with the user message they were sent with, so they show up in conversation history like chat attachments do.

On-screen content

During a voice call, some content is better shown on screen (the canvas) than spoken aloud: charts, tables, long lists, approval chains, code, formulas, or anything that reads awkwardly out loud. These settings, in the On-screen content section, control when the AI assistant pushes content to the canvas instead of speaking it directly. They apply to voice calls only (the Chat page links straight to this section); the Charts and Table settings on the Chat page do not apply here.

On-screen content section saying the Charts and Table settings on the Chat page do not apply to voice calls, with Let the assistant decide Off and an empty Always show on screen field, then the Sending what is on screen section with Send to WhatsApp and the button Go to Channels > WhatsApp

Sending what's on the canvas. Mid-call, a signed-in user can ask the assistant to send what is currently on the canvas to an email address, or simply "to my email", which uses the address of their account. With Send to WhatsApp on (in the Sending what is on screen section, right below), they can also send it to a WhatsApp number. The assistant reads the number or spells the address back and waits for a yes before sending. The recipient is told who sent it: the WhatsApp message and the shared page start with the user's name. The email comes from the agent's own -noreply address (for Atlas, atlas-noreply@qlar.ai), and the WhatsApp message from the agent's own WhatsApp number, so Send to WhatsApp needs a WhatsApp channel on the public or a private server. A picture arrives as an image; a chart or table arrives as a link that works for 7 days. Users who aren't signed in are asked to sign in first. If the organization has Privacy Mode on, sending is turned off.

Let the assistant decide

When on, the AI decides for itself, response by response, which parts to speak conversationally and which parts to show on the canvas instead, not limited to charts and tables. Leave this off if you'd rather rely only on the explicit list below.

Always show on screen

Never spoken. A list of free-text content types (for example chart, table, approval chain, mathematical formula, cards) that must always be rendered on the canvas instead of spoken as plain text, whenever a response contains that kind of content. This applies regardless of whether Let the assistant decide is on.

Pick a suggestion or type your own content type and press Enter; remove one with the Γ— on its chip. This only changes where the content is rendered: the agent still speaks whatever explanation the answer calls for alongside it.

Atlas leaves both settings empty here and describes what goes on screen in the immersive view instructions below instead.


Immersive view instructions

In the fold Immersive view options, the field What the assistant should do in the immersive view is free text added to the AI assistant's instructions only for calls the user is watching in the immersive view. A call shown as a transcript never receives them, because there the answer is already readable as text β€” instructions written for a screen with nothing to read ("show it, don't say it") would work against that.

Call screen section with the fold Immersive view options open: Opening sequence with Play the opening sequence Off, Subtitles with User transcription and Assistant subtitles both Off, Progress indicator with Position Top left, and Immersive view instructions with the field What the assistant should do in the immersive view containing: Show approval chains as a numbered list card. Show project status (KG codes) as a short card with status, stage and progress. Speak only the headline; keep details on screen.

Write it as instructions to the assistant, not as text the user sees. Atlas uses:

Show approval chains as a numbered list card. Show project status (KG codes) as a short card with status, stage and progress. Speak only the headline; keep details on screen.

With this, a question such as "Who has to approve a project worth $2.5 million?" gets a short spoken headline, while the three approvers appear on screen as a numbered list.

Where these instructions conflict with the On-screen content settings above, these win, because they are added last. Leave the field blank to add nothing at all.

Because Voice screen style is only a starting point, an agent set to Transcript can still have users who switch to the immersive view themselves, and they will get these instructions; the page says so when that is your configured default.

In Natural conversation, these instructions also guide the Answering AI, which prepares what's shown on screen. The voice itself receives only roughly the first 4,000 characters, and the page warns when the text is longer.


Subtitles

The immersive view has no transcript, but it can show what is being said as subtitles at the bottom of the screen. The Subtitles card in the fold Immersive view options has two switches, both Off by default, so you can turn on either one or both:

SwitchWhat the user sees
User transcriptionThe user's own words, in light grey, appearing as they speak.
Assistant subtitlesThe assistant's words, in white, revealed in step with its voice rather than all at once.

The subtitles sit on a dark glass panel over the end call, microphone and attach buttons while someone is speaking. A moment after the speaking stops, the panel fades out and the buttons can be used again. A user who wants the buttons back sooner can tap the subtitles to hide them until the next exchange.

When an answer is interrupted, its subtitle stops at the last word the user actually heard. Long answers keep only their last two lines on screen.

Atlas turns on both, so site supervisors calling from a noisy yard can check what they said and read Atlas's answer.

Subtitles apply only to the immersive view: the transcript view always shows the conversation as text. As with the instructions above, an agent set to Transcript shows them to users who switch to the immersive view themselves.


Progress indicator

While the assistant looks something up or prepares a card, the immersive view says what it is doing ("Checking the approval rules…") next to a spinner. The Progress indicator card in the fold Immersive view options decides where that appears, in its Position field:

OptionWhat the user sees
Top left (default)A short line under the assistant's name in the top bar, with a spinner beside the name.
Center of the screenA rounded, translucent box over the middle of the screen with the spinner and the line. The box is drawn in the opposite colours of the call screen (dark on a light screen, light on a dark one) so it stands out over the empty call screen and over any card, and a sheen sweeps across the text from left to right while the assistant works.

With Center of the screen, the top bar no longer shows the progress, so there is one place to look. The box appears only while the assistant is working and fades away a moment after it finishes, once the line has had time to be read. Users who have asked their device for reduced motion see the text without the sheen.

Atlas uses Center of the screen, so a supervisor glancing at a phone on site can tell straight away that Atlas is still looking something up.

Like the subtitles, this applies only to the immersive view.


Opening sequence

The Opening sequence card in the fold Immersive view options makes a new immersive call open with a small show instead of a quiet screen:

  1. A short sound plays (about four seconds) while the logo glows.
  2. After a short pause, the assistant reads your introduction aloud in its own voice.
  3. The logo rises, and the assistant says a short lead-in ("For example, you could ask:") in its language.
  4. Your example messages appear under the logo one at a time, each fading in just as the assistant starts saying it.
  5. The assistant says your closing line, and it appears on screen under the examples. Unless you changed it on the Introduction page, it is "What can I help you with?", in the language your introduction is written in (which may differ from the Dialog language); if there are no examples and your introduction already ends with a question, this default is left out. Clear the closing line there for none.

When it is over, the examples and the closing line stay on screen. Users can tap an example to ask it.

SettingWhat it does
Play the opening sequenceTurns the sequence on for new immersive calls. Off (the default): calls start quietly, as before.
Play opening soundOn: the sound plays first. Off: the music is skipped and the assistant starts speaking after a short pause.
Opening soundOne of ten sounds, from Cinematic Rise to Gamelan Welcome. Use the play button beside it to hear the sound you picked before you save.
Read the introduction aloudOn: the assistant says your introduction message first. Off: the introduction is skipped and it goes straight to the example messages.

The words come from the Introduction page: the introduction message (its variables filled in for the caller, its buttons left out), then as many example messages as the chat shows. With neither, only the sound plays.

Callers are never held up by it:

  • Saying a clear question or remark at any point stops the sequence, and the assistant answers right away. If what the caller said is unclear or cut off, the assistant says the closing line ("What can I help you with?" by default) instead, or briefly asks what they would like help with when there is none.
  • A short "hello" back, an "mm-hm" or the assistant's own voice coming back through the speaker does not stop it.
  • Tapping an example stops the sequence and asks that example.
  • A call that reconnects after a dropped connection, or one shown as a transcript, starts without it.

Atlas plays Cinematic Rise, reads "Hi, I'm Atlas. Ask me about approvals, projects or site safety.", then shows "Who has to approve a project worth $2.5 million?" and two more examples, and closes with "What can I help you with?".

On Natural conversation, the assistant reads the sequence in one go and the examples appear as it reaches each one. A very long introduction is shortened at a sentence end, and the last examples may be left out, so that what appears on screen is what is said.

The spoken part uses credits like any other speech from the assistant: as voice tokens on Turn-by-turn, and as call time on Natural conversation. It does not count as a message.


Check the result

  1. Click Save all changes.
  2. Start a voice conversation in the preview (+ menu next to the chat input). The call should open in the voice screen style you chose. In the immersive view, the agent's logo is shown on an otherwise empty screen until something appears on the canvas.
  3. Ask for something that belongs on screen. For Atlas, say "Who has to approve a project worth $2.5 million?". The approval chain should appear as a numbered list card on the canvas while Atlas speaks only a short headline.
  4. If attachments are allowed, attach an image or document and ask a question about it.
  5. If you turned on Subtitles, check that your words and the assistant's answer appear at the bottom of the screen, and that the panel fades out after the answer so the call buttons come back.
  6. If you chose Center of the screen, ask something that needs a lookup and check that the box with the spinner appears in the middle of the screen and fades away after the answer starts.
  7. If you turned on the Opening sequence, start a new immersive call: the sound should play, the assistant should read your introduction, your examples should appear one at a time as they are said, and the assistant should finish with your closing line ("What can I help you with?" by default), shown under the examples. Start another call and say a question while the introduction is being read: the sequence should stop and the assistant should answer.

Next steps