2.5. Voice Input and Read Aloud
Overview
Instead of the keyboard, you can speak your question, and you can listen to the AI agent's answer as sound. What you say is turned into text and placed in the message box, and you check it and send it yourself. This chapter explains how to enter text with the microphone, checking the transcribed text, having an answer read aloud, and choosing the reading voice. Use it to ask a question where typing is hard, or when you want to listen to a long answer instead of reading it.
Permission: General users. The buttons appear only when the feature is turned on in your installation.
Requirements
- "Voice input" and "Read aloud" appear only when the feature is turned on in your installation.
- "Voice input" is shown only when you connect over a secure connection (https) in a browser that can use a microphone.
- Your AI agent workspace must be connected. If it is not connected, "Voice input is unavailable because your AI Agent Workspace is not connected" is shown.
If there is no microphone button, check in this order.
- Check whether the browser address starts with
https://. - If it starts with
http://, ask your administrator for a secure connection address. - If the address is correct but there is still no button, ask your administrator to turn on the voice input feature. Example request: "The voice input (microphone) button does not appear in AAS chat. Please turn the feature on."
Using Voice Input
"Voice input" is in the message box of the chat screen and the chat pop-up.
Figure 2.5-1. The message box with the "Voice input" button
- Select the "Voice input" (microphone) button in the message box.
- If the browser asks to use the microphone, choose "Allow."
- Say your question.
- When you finish speaking, select the "Finish recording" (check mark) button.
If this works, the row below the message box changes to a recording row, and the elapsed time is shown like "0:12 / 1:00." If you refused the microphone or no sound comes in, the message in the "Allowing the Microphone the First Time" table below appears.
| Time | Screen |
|---|---|
| 50 seconds | "10 seconds left" is shown. |
| 1 minute | "Recording stopped at 1 minute" is shown and recording ends by itself. One recording can last up to 1 minute. |
To discard a recording, select the "Cancel recording" (close mark) button or press Esc. After recording ends, while it is being turned into text, "Transcribing…" is shown, and you can stop it with "Cancel transcription."
Where Your Voice Goes
When you finish recording, your voice is sent to the AAS server. The server passes it through the AI agent workspace to the speech recognition model your installation connected, and turns it into text. The sound is not stored on the server, and it does not remain after it is turned into text. The usage records do not keep the text of what you said either. If transcription fails, the last recording is kept briefly in your browser's memory only, so you can try again. It is discarded when you start a new recording or switch chats.
Allowing the Microphone the First Time
The first time, the browser shows a window asking whether to use the microphone. Choose "Allow." If you did not allow it, the following messages appear.
| Message | What to do |
|---|---|
| Allow the microphone in the browser prompt | In the browser's microphone window, choose "Allow." |
| Allow the microphone from the site information icon at the left of the address bar | You already refused. In the site information at the left of the address bar, change the microphone to allowed. |
| No microphone found | Check that a microphone is connected to the computer. |
| Another program is using the microphone | Close the other program that uses the microphone. |
| No sound detected. Check your microphone | Check whether the microphone is muted and which input device is selected. |
Checking the Transcribed Text and Sending
When transcription finishes, "Transcribed text added to the message box. Check it and press Enter to send" is shown, and the text goes into the message box. The text goes where the cursor was when you started recording. Text entered by voice is not sent automatically.
- Read the text in the message box.
- Correct any wrong characters.
- Press
Enterto send.
You cannot send before transcription finishes, and "You can send after transcription finishes" is shown.
When Transcription Does Not Work
| Message | What to do |
|---|---|
| No speech recognized | Speak again in a quiet place. |
| Could not transcribe. Select retry | Select "Retry" on the recording row. To discard the recording, select "Discard recording." |
| Transcription is busy right now. Select retry | Select "Retry" in a moment. |
| The recording is too long. Speak in shorter parts | Split what you say and enter it in several turns. |
| The AI Agent Workspace connection was lost. Select retry once it reconnects | After the connection returns, select "Retry." The recording stays briefly in your browser in the meantime. |
| Recording in another window | You can record in only one message box at a time. Finish the recording in the other window and select again. |
If you switch chats, open a new chat, or move to another screen, a recording in progress is discarded.
Listening to an Answer
When the AI agent's answer is finished, a "Read aloud" button appears below the answer.
- Select "Read aloud" below the answer.
- "Getting ready to read" is shown, and when it is ready, sound plays.
- To stop while it is reading, select "Stop reading."
The reading range is about the first 15 minutes of reading time. For a longer answer, "Reading only the first 15 minutes or so" is shown. Read aloud sends the answer's text to the AAS server to turn it into sound. The sound is not stored on the server. In the following cases, the button cannot be selected.
| Situation | Message |
|---|---|
| While recording | Cannot read aloud while recording |
| While the answer is being written | Available once the answer is finished |
If there is no text to read, "Nothing to read" is shown. If it cannot read, "Could not read this aloud" is shown. If requests are crowded, "Too many read requests right now" is shown. In these cases, select "Try again." If you move to another chat or close the chat window while it is reading, the reading stops. Reading continues if the browser tab is hidden.
Choosing the Reading Voice
- In the left menu, open "Settings › My display."
- In the "Read-aloud voice" field, choose a voice. After each voice name you see Female or Male.
- Select "Preview" to hear the sound. If you do not like it, switch to another voice and listen again.
Figure 2.5-2. The "Read-aloud voice" field
While a preview plays, the button changes to "Stop preview." The value you choose is saved only in this browser. In another browser or on another computer, you must choose again. If you do not choose a voice, the first voice in the list is used. The "Read-aloud voice" field appears only when the read aloud feature is turned on.