AVS UX Attention System

Alexa is a single Amazon-branded personality that is coherent and familiar to customers across all products. While the physical products might be different, the attention system ensures Alexa behaves predictably and with familiarity. This consistency creates customer trust and strengthens the customer’s understanding of Alexa.

Alexa’s attention system is comprised of non-verbal audio and visual components that work together to communicate all of Alexa’s different states to the customer. Color, sound, and animation are critical for effectively communicating state since there is no other visual or verbal explanation of the attention system on your product. Audio and visual cues must be synced so that Alexa’s state change indicators occur simultaneously as the customer wakes, speaks to, and listens to Alexa.

• Attention System Components
  • Alexa Sound Library
  • Alexa Attention State Colors
  • Chrome and LED Placement
• Attention System States
  • Idle
  • Core
  • Microphone Off
  • Alerts
  • Notifications
  • Pairing & Discovery
  • Errors
  • Device States
• Attention System Implementations
  • Multicolor LEDs
  • Voice Chrome
  • Sound Cues Only

Attention System Components

Different product types, shapes, and features will warrant different visual implementations of the attention system. Although we have written guidance for many different implementations, they are not all equal in their ability to effectively communicate.

To properly convey all of the different states to our customers, multiple multicolor LEDs are strongly recommended for non-screen products. Customers otherwise will be unable to recognize some Alexa behaviors and features. At a minimum, your device must clearly indicate when it is streaming audio to the cloud.

Alexa Sound Library

Alexa audio cues are short sounds that play to indicate an Alexa state change. They can be used alone or in concert with visual cues provided by voice chrome or LEDs. The fewer visual cues your device has, the more important the audio cues are. Even for a device with strong visual cues, there may be a chance the customer will not be looking at it while activating or speaking to Alexa, making the audio cues equally important as the visual cues. In all cases, audio cues should clarify rather than confuse.

Amazon provides an Alexa Sound Library, containing three different sound sets for developers. Each set is targeted to a different speaker quality range. Only sounds that must be stored on device are provided. All other sounds are part of the Alexa response.

The decision about which set to use depends upon your product's speaker capabilities and which files sound best on the product. The three sets are Full, Medium and Small, and each sound file is prefixed with a corresponding code: "ful", "med" and "sml".

The Sound Library can be downloaded from the Resources tab of the AVS Developer Portal.

Alexa Attention State Colors

Colors help indicate to customers what’s going on with the product. Each color represents a family of interactions. We use a system of colors and animations to communicate the state of the product, and how the customer may interact with it.

Chrome and LED Placement

In order for a visual attention system to be effective, the customer has to be able to see it. Think about how the customer will use your product and what surfaces would be easiest to see for visual communication, for example whether can the visual cues be seen in every direction.

Devices like a sound bar should have LEDs placed on the customer-facing side of the product, the front, rather than on the top where the customer won't be able see it. Similarly for screen-based devices, keep in mind where your customer will be in relation to your product to ensure the chrome is placed where they can easily see it.

Attention System States

Attention states address the personality of Alexa at a high level across all domains.

Idle

Alexa is idle when she isn’t in another state. Idle is important as it sets the initial state of Alexa that must contrast with all other states. The easiest way to create that contrast visually is for idle to not use visuals (i.e. LEDs off) and all other states light up the LEDs or display on screen. This way the user will easily know when Alexa is actively trying to communicate versus passively waiting for a request.

Core

The Core Alexa states are: Listening, Thinking and Speaking. For products with visual cues, it is required that these three states and the Idle state are distinguishable from each other.

Listening

The Listening state starts when Alexa has been initiated via wake word, tap-to-talk, or push-to-talk, and the microphone begins streaming the customer’s request to the Alexa Voice Service.

There are three stages to the Listening state:

• Start Listening When the product’s microphone(s) have been activated, transitions into the Listening state and waits for a request from the customer.
• Active Listening When the customer begins speaking, Alexa transitions into an Active Listening state.
• End Listening When the customer's end of speech is identified, Alexa transitions out of Listening state.

The amount of visual feedback your product provides will determine the requirements for audio cues. Products without strong visual cues are required to use audio cues to indicate when Alexa has started and stopped listening. Products with a strong visual system have the option of setting these audio cues to off by default.

Notes:

• Voice-initiated products use a different start-listening sounds than touch-initiated products.
• Voice-initiated products use the same start-listening sound whether the customer uses touch or voice to initiate an Alexa interaction on the product.
• In multi-turn interactions on voice-initiated devices, the start-listening audio cue should only be played on the first turn of the interaction. In subsequent turns, the start-listening sound should not be played.
• In multi-turn interactions on touch-initiated products, the start-listening audio cue should be played every time Alexa starts listening and the end-listening audio cue must be played at the end of every turn.

Thinking

When a customer has completed a request, Alexa enters the Thinking state. This state lets the customer know the microphone is no longer active and Alexa will respond shortly. During the Thinking state, no additional voice input from the customer is accepted.

Note: No sounds should play when Alexa starts or stops Thinking.

Speaking

The Speaking state is displayed when Alexa is responding to a request with a TTS (text to speech). This state is not displayed when Alexa is responding with long running mixable media such as music, books and Flash Briefings.

Note: No sounds should play when Alexa starts or stops Speaking.

Microphone Off

On a voice-initiated implementation, the customer can block Alexa from waking by turning on the microphone-off mode. In this state, the product is not capable of detecting the Alexa wake word and will not send any customer utterances to Alexa. The microphones should be physically powered down, as well as any camera on the product. Alerts and Notifications are still played when the microphone is off.

Notes:

• Voice-initiated products are required to use audio cues when the customer turns microphones on and off. The microphone on/off sounds must always be enabled, with no option for the user to disable them.
• The microphone on/off sounds shall play in all instances, even during Alexa TTS responses.
• Voice-initiated product must visually display that microphones are off so that the user can easily understand why Alexa isn't responding to the wake word.
• Touch-initiated products do not have a Microphone Off state because their microphones are always off until the customer physically initiates an interaction.

Alerts

Customers can ask Alexa to alert them after a specified duration or at a specific time. Alexa supports Timers, Alarms and Reminders.

Important: Alerts must activate when they are scheduled, even if your product has lost internet connectivity.

When a Timer or Alarm activates, the device must play the appropriate sound and visual cues on a loop until the customer stops the alert or for an hour, whichever comes first. When a Reminder activates, the product must play the audio cue included in the Alexa response, and Alexa should then speak the Reminder. Information on the number of times the Reminder should be looped and the duration of the pause between loops is included in the Alexa response.

If Alexa is responding (e.g. giving the 10-day weather forecast) while an alert sound should be played, the short versions of the audio and visual cues should be played once every 10 seconds. Once the Alexa response is complete, the full length version of the alert should be played as described above.

If you implement Alerts interface version 1.3, customers can manage alert volumes and set custom alert sounds. In order to allow the customer to choose a custom sound, the device must accept the URL sent during alert confirmation. If a device stores only the default alert sound, the customer will not be able to change it.

Customers can also ask Alexa to delete alerts, for example "Alexa, delete the 1-hour timer." There are no UX indicators required other than Alexa's Core attention states during the conversation. If you've implemented the Alerts interface 1.3 or later, customers are able to delete all the alerts of a specified type, such as "Alexa, delete all alarms."

Notes:

• The customer must be able to stop an alert by some method other than voice so that they can stop the alert even if your product has lost internet connectivity.
• We recommend using your product's action button, or providing an on-screen control, to stop a playing an alert. Pressing the button or control should stop the alert playing but not wake Alexa.
• Alert sounds are sent with the alert confirmation, and must be stored on the product so that alerts can be played at the correct time even if your product loses internet connectivity. If your device supports Alerts interface version 1.3 or later, this is true of custom alert sounds as well.
• If your device allows the customer to change audio volume, there is no specific UX indicator required.

Timers and Alarms

Customers can ask Alexa to alert them after a specified duration by setting a Timer, or at a specific time by setting an Alarm. Timers and Alarms​ use distinct sounds, but otherwise, behave the same way.

Notes:

• All products are required to use audio cues to indicate when a Timer or Alarm goes off.
• Visual cues for Timers and Alarms are highly recommended.
• The Timer audio cue should be used only for Alexa Timers, and the Alarm audio cue should be used only for Alexa Alarms.
• As of Alerts interface 1.3, the customer can change alert volume separately from system volume (through app, buttons, or on-screen GUI). Note that changes to alert volume made while an alert is actively sounding are temporary, applying only to the active alert, and return to the previously set volume once the active alert is complete. Likewise, if there is content playing, such as music, and the system volume is higher than the alert volume, the alert plays at the system volume level. This is also a temporary volume change and the level returns to the custom volume setting when there is no content playing.

Reminders

Customers can ask Alexa to alert them with a TTS message after a specified duration or at a specific time by setting a Reminder.

An example Reminder output is “[audio cue] Here’s your reminder: Take out the garbage. [pause] [audio cue] Here’s your reminder: Take out the garbage.”

For more about implementing Reminders, see ​Enable Named Timers and Reminders and the Alerts Overview​.

Notes:

• The URLs for the audio cue file and the TTS file are included in the Alexa response.
• The Reminder files must be stored locally on the device so that the Reminder can be played at the correct time even if your product loses internet connectivity.
• Reminder audio must be played at the system volume level.

Notifications

Notification indicators inform customers that new content is available from Alexa Skills. The customer has to request that Alexa notify them about an interesting or important event.

Notes:

• When a Notification arrives, the new-notification sound must be played immediately.
• Even if Alexa is responding (e.g. reading a Flash Briefing, or giving the 10-day weather forecast) or media is playing when a notification arrives, the audio cue should be played immediately.
• The audio file should only be played once for each new Notification and should not be played on a loop.
• The URL for the audio file is included in the Alexa response.
• In case the downloadable file is unavailable, the new-notification sound should be stored locally so that it can be played as a fall-back.
• The visual indicator that Notifications are waiting to be heard should be displayed during the device's Idle state until the customer has listened to all notifications.

Pairing & Discovery

Bluetooth

Customers can pair devices such as mobile phones and Bluetooth-enabled speakers to stream music and control media playback.

Notes:

• The Alexa Bluetooth-connected and Bluetooth-disconnected sounds should be played when the product receives a connect or disconnect success event.
• The sounds should be played whether the customer uses a voice command or a companion app to manage the connection.

Errors

When errors that prevent Alexa from functioning, for example, Wi-Fi is not available, occur, they are presented visually to the customer through Alexa’s attention system. System errors are only displayed after a user request has failed. For example, if a product loses connection to Wi-Fi, the error would only be displayed if the user makes a request, not the moment the device lost its connection.

Notes:

• Product sounds, and not Alexa sounds, should be used during setup or if there is an issue with connectivity.
• Error sounds should not be added to an Alexa response. This is true even if the customer’s speech was not understood by the service.

Device States

In addition to the Alexa attention states, your product may need to display other states. We do not provide recommendations for those other states, but they must be distinct from the Alexa attention states, and cannot use Alexa audio cues.

Device Setup

On Amazon Alexa products, blue and cyan are used to show that the device is booting up, or during the initial Out of Box Experience (OOBE).

Device Changes

On Amazon Alexa products, white is used to show that the device state is changed in some way, such as volume changed or muted, or powered off.

Configuration Changes

On Amazon Alexa products, orange is used to show that the configuration changes are being made on the device, such as Wi-Fi configuration, wake word, or factory reset.

Attention System Implementations

Different products will work better with different implementations of the the Alexa Attention System. Here we describe the most common ones.

For screen-less devices, we strongly recommend using multiple color LEDs when possible to communicate the Alexa states to your customer. For screen-based devices, we require you to use the colors provided above when communicating Alexa states.

The sections below provide guidance for these implementations:

• Multicolor LEDs (radial array, linear array and single LED)
• Voice Chrome
• Sound Only

Multicolor LEDs

• Multicolor Radial
• Multicolor Linear
• Multicolor Single

We highly recommend this implementation as it is most similar to the visual layout of the Amazon Echo. Your product will be able to clearly and consistently display all the Alexa attention states using this implementation.

State	Details	Visuals	Sounds
Idle	No Alexa audio or visual cues when the product is idle.		None
Listening Start	Visual and audio cues play once when the customer wakes Alexa by voice or touch and the microphone becomes active. Different start listening audio cues are used for voice-initiated and touch-initiated products. For devices with this level of visual cues, the start listening audio cue can default to off. However, we recommend enabling the sound by default and allowing the user to disable the sound in your companion app.		Voice-initiated Touch-initiated
Listening Active	Visual cue persists while Alexa is capturing speech from the customer. Single LED segment should point in the direction of the customer's voice while the microphone is active.		None
Listening End	Audio cue, but not visual cue, is played if Alexa transitions into the Thinking state because the end of a customer utterance is identified. Visual cue, but not audio cue, plays if Alexa stops Listening and does not go into the Thinking state but instead returns to Idle. This can occur for a timeout or manual request to end the Alexa interaction. For devices with this level of visual cues, the end listening audio cue can default to off. However, we recommend enabling the sound by default and allowing the user to disable the sound in your companion app.
Thinking	Visual cue plays in a loop while Alexa is processing, or 'thinking about' what the customer has said. Displaying this state ensures that the customer understands that the interaction has not ended.		None
Speaking	Visual cue plays in a loop while Alexa is responding to the customer and TTS is playing from device.		None
End Speaking	Visual cue plays once when Alexa has finished responding to the customer and returns to Idle.		None
Mic On to Mic Off	Visual and audio cue must play once when the customer turns off the microphone.
Mic Off	Visual cue persistently communicates the Mic Off state to the customer.		None
Mic Off to Mic On	Visual and audio cue must play once when the customer turns on the microphone.
Timer	Visual and audio cues play in a loop when a Timer comes due and Alexa is not currently responding with speech. Looping of the visual and the audio cue should continue until the customer stops the Timer, or for one hour, whichever comes first.
Timer Short	Short Timer audio and visual cues play once every 10 seconds when a Timer comes due while Alexa is responding with speech. Once the Alexa speech response ends, the full Timer response should be played as described above.
Alarm	Visual and audio cues play in a loop when an Alarm comes due and Alexa is not currently responding with speech. Looping should continue until the customer stops the Alarm, or for one hour, whichever comes first.
Alarm Short	Short Alarm audio and visual cues play once every 10 seconds when an Alarm comes due while Alexa is responding with speech. Once the Alexa speech response ends, the full Alarm response should be played as described above.
Reminder	Plays when a Reminder comes due and Alexa is not currently responding with speech. The visual cues and audio files should be played in a loop as defined in the directive.		Provided by service
Reminder Short	Plays when a Reminder comes due while Alexa is responding with speech. Audio cue (but not TTS) and visual cue must be played once every 10 seconds. Once the Alexa speech response ends, the full Reminder response should be played as described above.		Provided by service
Incoming Notification	Visual and audio cues play once on delivery of a Notification. The audio cue must be played.
Queued Notification	Visual cue loops when a customer has a queue of unheard Notifications until the customer clears the Notification queue.		None
Bluetooth Connected	Visual cue is flashed three times and audio cue is played once when a Bluetooth connection is made.
Bluetooth Disconnected	Visual cue flashes three times and audio cue is played once when a Bluetooth connection is ended.
Error	Visual cue flashes three times when an error is encountered.		None

This implementation can clearly and consistently display all the Alexa attention states as well as the multicolor radial LED arrangement, depending upon the size and location of the LED bar. If the LED bar is small or difficult for a customer to see, the start and end listening sound cues must be enabled by default.

State	Details	Visuals	Sounds
Idle	No Alexa audio or visual cues when the product is idle.		None
Listening Start	Visual and audio cues play once when the customer wakes Alexa by voice or touch and the microphone becomes active. Different start listening audio cues are used for voice-initiated and touch-initiated products. Depending upon the size and location of your LEDs, the start listening sound might be able to default to off. However, we recommend enabling the sound by default and allowing the user to disable the sound in your companion app.		Voice-initiated Touch-initiated
Listening Active	Visual cue persists while Alexa is capturing speech from the customer. Single LED segment should point in the direction of the customer's voice while the microphone is active.		None
Listening End	Audio cue, but not visual cue, is played if Alexa transitions into the Thinking state because the end of a customer utterance is identified. Visual cue, but not audio cue, plays if Alexa stops Listening and does not go into the Thinking state but instead returns to Idle. This can occur for a timeout or manual request to end the Alexa interaction. Depending upon the size and location of your LEDs, the end listening sound might be able to default to off. However, we recommend enabling the sound by default and allowing the user to disable the sound in your companion app.
Thinking	Visual cue plays in a loop while Alexa is processing, or 'thinking about' what the customer has said. Displaying this state ensures that the customer understands that the interaction has not ended.		None
Speaking	Visual cue plays in a loop while Alexa is responding to the customer and TTS is playing from device.		None
End Speaking	Visual cue plays once when Alexa has finished responding to the customer and returns to Idle.		None
Mic On to Mic Off	Visual and audio cue must play once when the customer turns off the microphone.
Mic Off	Visual cue persistently communicates the Mic Off state to the customer.		None
Mic Off to Mic On	Visual and audio cue must play once when the customer turns on the microphone.
Timer	Visual and audio cues play in a loop when a Timer comes due and Alexa is not currently responding with speech. Looping of the visual and the audio cue should continue until the customer stops the Timer, or for one hour, whichever comes first.
Timer Short	Short Timer audio and visual cues play once every 10 seconds when a Timer comes due while Alexa is responding with speech. Once the Alexa speech response ends, the full Timer response should be played as described above.
Alarm	Visual and audio cues play in a loop when an Alarm comes due and Alexa is not currently responding with speech. Looping should continue until the customer stops the Alarm, or for one hour, whichever comes first.
Alarm Short	Short Alarm audio and visual cues play once every 10 seconds when an Alarm comes due while Alexa is responding with speech. Once the Alexa speech response ends, the full Alarm response should be played as described above.
Reminder	Plays when a Reminder comes due and Alexa is not currently responding with speech. The visual cues and audio files should be played in a loop as defined in the directive.		Provided by service
Reminder Short	Plays when a Reminder comes due while Alexa is responding with speech. Audio cue (but not TTS) and visual cue must be played once every 10 seconds. Once the Alexa speech response ends, the full Reminder response should be played as described above.		Provided by service
Incoming Notification	Visual and audio cues play once on delivery of a Notification. The audio cue must be played.
Queued Notification	Visual cue loops when a customer has a queue of unheard Notifications until the customer clears the Notification queue.		None
Bluetooth Connected	Visual cue is flashed three times and audio cue is played once when a Bluetooth connection is made.
Bluetooth Disconnected	Visual cue flashes three times and audio cue is played once when a Bluetooth connection is ended.
Error	Visual cue flashes three times when an error is encountered.		None

As it is difficult to clearly and consistently display all the Alexa attention states with a single LED, we recommend using multiple LEDs instead. The start and stop listening sound cues must be enabled by default for these devices.

State	Details	Visuals	Sounds
Idle	No Alexa audio or visual cues when the product is idle.		None
Listening Start	Visual and audio cues play once when the customer wakes Alexa by voice or touch and the microphone becomes active. Different start listening audio cues are used for voice-initiated and touch-initiated products. The start listening audio cue must be enabled by default, and we discourage you from allowing the user to disable the sound in your companion app.		Voice-initiated Touch-initiated
Listening Active	Visual cue persists while Alexa is capturing speech from the customer.		None
Listening End	Audio cue, but not visual cue, is played if Alexa transitions into the Thinking state because the end of a customer utterance is identified. Visual cue, but not audio cue, plays if Alexa stops Listening and does not go into the Thinking state but instead returns to Idle. This can occur for a timeout or manual request to end the Alexa interaction. The end listening audio cue must be enabled by default, and we discourage you from allowing the user to disable the sound in your companion app.
Thinking	Visual cue plays in a loop while Alexa is processing, or 'thinking about' what the customer has said. Displaying this state ensures that the customer understands that the interaction has not ended.		None
Speaking	Visual cue plays in a loop while Alexa is responding to the customer and TTS is playing from device.		None
End Speaking	Visual cue plays once when Alexa has finished responding to the customer and returns to Idle.		None
Mic On to Mic Off	Visual and audio cue must play once when the customer turns off the microphone.
Mic Off	Visual cue persistently communicates the Mic Off state to the customer.		None
Mic Off to Mic On	Visual and audio cue must play once when the customer turns on the microphone.
Timer	Visual and audio cues play in a loop when a Timer comes due and Alexa is not currently responding with speech. Looping of the visual and the audio cue should continue until the customer stops the Timer, or for one hour, whichever comes first.
Timer Short	Short Timer audio and visual cues play once every 10 seconds when a Timer comes due while Alexa is responding with speech. Once the Alexa speech response ends, the full Timer response should be played as described above.
Alarm	Visual and audio cues play in a loop when an Alarm comes due and Alexa is not currently responding with speech. Looping should continue until the customer stops the Alarm, or for one hour, whichever comes first.
Alarm Short	Short Alarm audio and visual cues play once every 10 seconds when an Alarm comes due while Alexa is responding with speech. Once the Alexa speech response ends, the full Alarm response should be played as described above.
Reminder	Plays when a Reminder comes due and Alexa is not currently responding with speech. The visual cues and audio files should be played in a loop as defined in the directive.		Provided by service
Reminder Short	Plays when a Reminder comes due while Alexa is responding with speech. Audio cue (but not TTS) and visual cue must be played once every 10 seconds. Once the Alexa speech response ends, the full Reminder response should be played as described above.		Provided by service
Incoming Notification	Visual and audio cues play once on delivery of a Notification. The audio cue must be played.
Queued Notification	Visual cue loops when a customer has a queue of unheard Notifications until the customer clears the Notification queue.		None
Bluetooth Connected	Visual cue is flashed three times and audio cue is played once when a Bluetooth connection is made.
Bluetooth Disconnected	Visual cue flashes three times and audio cue is played once when a Bluetooth connection is ended.
Error	Visual cue flashes three times when an error is encountered.		None

Voice Chrome

Voice Chrome should be used on screen-based products to indicate the Alexa attention states. If your product supports Display Cards as well, see Display Card guidance on how cards and chrome should be combined. For those states not explicitly documented below, refer to the LED linear images above for guidance.

State	Details	Sounds
Listening Start	Visual and audio cues play once when the customer wakes Alexa by voice or touch and the microphone becomes active. Different start listening audio cues are used for voice-initiated and touch-initiated products.	Voice-initiated Touch-initiated
Listening Active	Visual cue persists while Alexa is capturing speech from the customer.	None
Thinking	Visual cue plays in a loop while Alexa is processing, or 'thinking about' what the customer has said. Displaying this state ensures that the customer understands that the interaction has not ended.	None
Speaking	Visual cue plays in a loop while Alexa is responding to the customer and TTS is playing from device.	None
Mic On to Mic Off	Visual and audio cue must play once when the customer turns off the microphone.
Mic Off	Visual cue persistently communicates the Mic Off state to the customer.	None
Mic Off to Mic On	Visual and audio cue must play once when the customer turns on the microphone.

Sound Cues Only

We strongly recommend that products incorporate both audio and visual components into their attention systems, though we understand that some products are unable to support a strong visual attention system. Those products must rely on the sound cues to communicate the current state of Alexa, but will lack the ability to unambiguously convey some Alexa states. We strongly discourage a sound-only attention system for voice-initiated products unless they are able to clearly and persistently display the Microphone Off state.