What is the difference between Alexa Built-in and Alexa Connect Kit (ACK)?

Alexa Built-in covers products that embed the AVS Device SDK and run a full Alexa client on the device: wake-word detection, streaming to the AVS cloud, rendering of audio and visual responses. ACK targets simpler IoT devices whose Alexa logic runs on an Amazon-hosted firmware paired with a small Amazon module: the accessory exposes capabilities and Amazon carries the cloud integration. Functional-certification depth is significantly different between the two programmes, and the acoustic test plan applies in the Alexa Built-in case, not in a typical ACK setup where the device has no far-field microphone array.

Is the Alexa acoustic test plan replaceable by an EN 50332-type measurement?

No. EN 50332 covers maximum acoustic output for hearing protection on listening devices, which is a safety angle. The Alexa acoustic test plan covers the acoustic front end of the device as a voice-recognition input: wake-word detection rate, far-field SNR, AEC during music playback, noise suppression in diffuse fields, multi-talker resilience. The two test plans are orthogonal: a product must address both when it falls in scope, EN 50332 for the listening side, the Alexa plan for the microphone-input side.

Which DRM level does Google Cast require for premium video content?

For HD and UHD premium content distributed by services such as Netflix or Disney+, Cast video receivers must implement Widevine L1, the hardware-backed DRM level: keys are handled in a trusted execution environment, content decoding runs in a protected path. Widevine L3, software-only, is acceptable only for lower-resolution content or specific configurations. For Cast audio-only products, the DRM constraint is lower because the audio chain does not carry equivalent content protection.

What is the role of mDNS-SD and DIAL in Cast discovery?

mDNS-SD (Multicast DNS Service Discovery) is the local protocol the sender device, phone or browser, uses to discover Cast receivers on the same network. DIAL (Discovery And Launch) is the protocol that, after discovery, lets the sender launch a target application on the receiver. The Cast Application Framework sits on top of these two layers and manages the session, the messages exchanged between sender and receiver, and the synchronisation of state. Misbehaviour in mDNS-SD or DIAL is a common cause of pairing failure during certification.

Does AVS or Cast certification replace FCC and CE radio certification?

No. AVS and Cast certifications are programme certifications, cloud connection plus functional plus acoustic, run by Amazon and Google respectively. Wi-Fi, Bluetooth and any other radio interface must in addition be certified under the applicable regulatory regimes: FCC Part 15 in the United States, RED in Europe, MIC in Japan, KC in Korea, and so on. These layers stack on top of each other: a smart speaker for the European market needs CE marking under RED, Wi-Fi Alliance Wi-Fi certification, Bluetooth SIG qualification where Bluetooth is used, and the Alexa Built-in or Cast certification programme.

Is a physical mute button mandatory on Alexa Built-in products?

Yes for products with always-listening microphones. The Amazon Privacy and Security Requirements for Alexa Built-in require a hardware mute control that physically disconnects or visibly inhibits the microphone array, with an always-on LED indicator when the microphone is actively streaming audio to the cloud. The intent is to give the user a clear and verifiable signal of microphone state. Devices that only soft-mute by software are not compliant: pre-cert reviews catch this pattern frequently.

How does Matter casting interact with the existing Cast SDK?

Matter casting is an extension of the Matter specification that introduces casting roles, Casting Source and Casting Target, over the Matter fabric. The Cast SDK and Matter casting are not interchangeable: a Cast-only receiver does not become a Matter casting target automatically. Integrating both requires combining the Matter SDK with the Cast SDK on the device. The 2024 to 2026 horizon is one of progressive convergence, the timing of which depends on Matter version revisions and the ecosystem actors involved. The Cast over Matter angle is still being structured at the time of writing.

What are the most frequent pitfalls?

Acoustic front end with insufficient microphone SNR, leading to a wake-word detection rate below the threshold required by the Alexa acoustic test plan. AEC failing in the presence of device-driven music playback, masking the wake-word and blocking the dialogue. OAuth refresh-token rotation broken after a device reboot, which logs the user out and breaks cloud connection. Cast video frame drops at Widevine handshake, often the result of a key-rotation delay or a timing problem in the protected pipeline. Microphone mute LED not always lit while the mic is streaming, a privacy-rule breach. Factory reset that removes Cast keys without re-provisioning, leaving the device unusable as a Cast receiver after the reset.

Amazon AVS and Google Cast accessory certification

Author Hugues Orgitello Last updated 27 May 2026 ~14 min read

Guide - AVS, Alexa Built-in, Cast, Chromecast built-in

Integrating Amazon Alexa or Google Cast into a smart speaker, a connected display, a TV or a home accessory adds a programme-certification layer on top of regulatory radio and EMC certification. Amazon structures its programme around three tracks, Alexa Built-in with the AVS Device SDK, Alexa Connect Kit (ACK) for simpler IoT devices, and Alexa Built-in Lite for limited integrations. Google structures its programme around Chromecast built-in for audio, Chromecast built-in for TVs, and Google Assistant built-in, with extensions emerging around Matter casting. Both programmes share the same constraint families: cloud-connection conformity, acoustic front end and far-field microphone test plan for voice products, OAuth and credential management, privacy and microphone-indicator rules, DRM for video. This page presents the institutional map of the two programmes, the AFE and acoustic test-plan logic, the wake-word and DSP IP layer, the mDNS-SD and DIAL discovery stack on the Cast side, the Widevine track for video, and the cross-programme constraints with regulatory radio certification.

Institutional map

The two programmes are run by Amazon and Google as private programmes, separate from public regulatory regimes. The certificate they issue is contractual: it authorises use of the Alexa Built-in or Chromecast built-in brand and access to the corresponding cloud, under the terms of the developer agreement.

Actor	Scope	Type of decision
Amazon Developer	Enrolment, SDK access, Alexa cloud account management	Account creation, contractual access to AVS or ACK
Amazon Alexa certification team	Functional, acoustic and privacy review for Alexa Built-in	Acoustic and functional test-plan pass, brand authorisation
Google Developers	Cast SDK and Assistant SDK access, account management	Account creation, contractual access to the Cast cloud
Google certification team	Functional, acoustic and DRM review for Cast and Assistant	Cast or Assistant certification pass, brand authorisation
Accredited acoustic laboratories	Anechoic chamber, diffuse field, multi-talker conditions	Test reports compliant with the Amazon or Google test plan
Widevine and DRM partners	Hardware Widevine L1 integration, KeyBox provisioning	Integration certificate for premium video

The procedural counterpart for the manufacturer is, at Amazon, the Alexa Built-in or ACK certification team accessed through the Amazon Developer portal, and at Google, the Cast or Assistant certification team accessed through Google Developers. Direct enrolment without a developer account is not possible.

Amazon AVS: Alexa Built-in, ACK, Alexa Built-in Lite

The Amazon Alexa accessory programme comprises three certification tracks, defined by the functional depth of the Alexa integration and the location of the cloud logic.

Alexa Built-in

Alexa Built-in covers products that embed the full Alexa client on the device. The technical foundation is the AVS Device SDK (C++ on Linux or RTOS, Java derivatives on Android-based platforms) or the Alexa Voice Service Integration Kit (AVS-IK) for accelerated integration on partner SoCs. The device handles:

wake-word detection locally, by an embedded engine (Sensory TrulyHandsfree, Amazon proprietary engine, or equivalent),
streaming to AVS of audio captured after wake-word, over HTTP/2 with OAuth authentication through Login With Amazon,
rendering of audio responses, with possibly a graphical interface on a display,
device-side capabilities declared via Alexa Smart Home or Alexa Skill API.

This track imposes the highest level of functional and acoustic conformance.

Alexa Connect Kit (ACK)

ACK targets simpler IoT devices: connected appliances, plugs, sensors, kitchen accessories. The Alexa client logic does not run on the device. The product integrates a small Amazon-supplied module (Wi-Fi plus MCU) preloaded with Amazon-hosted firmware. The device exposes its capabilities through an Amazon API, and Amazon manages the cloud part, OTA updates and account integration.

This track has no acoustic test plan, because the device typically has no far-field microphone array. Functional certification covers integration with the Amazon API and conformance to the Amazon Privacy and Security Requirements for connected devices.

Alexa Built-in Lite

Lite covers an intermediate scope: limited Alexa integration on devices that have a microphone but not a far-field array, typical of headphones or wearable accessories. The acoustic test plan is reduced compared with the full Built-in version, and the supported Alexa feature set is more limited.

Track	Cloud logic	Microphone	Test-plan depth
Alexa Built-in	On device, AVS Device SDK	Far-field array typical	Full acoustic and functional test plan
ACK	Amazon-hosted firmware on dedicated module	Typically none	Functional only, no acoustic
Alexa Built-in Lite	On device, reduced scope	Mono or close-talk	Reduced acoustic test plan

Acoustic test plan and Acoustic Front End

Functional certification for Alexa Built-in rests on an acoustic test plan, the Alexa Acoustic Test Plan, complemented by the Alexa Functional Test Plan. The objective is to validate that the device's acoustic front end (AFE) lets the wake-word engine and the AVS-streamed dialogue work in realistic listening conditions.

Acoustic Front End

The AFE is the chain that picks up the user's voice, conditions it, applies acoustic echo cancellation when the device itself plays audio, and supplies a clean signal to the wake-word engine and the cloud. Its components:

far-field microphone array (typically 2 to 7 microphones depending on the device class), beamforming-capable,
AEC (Acoustic Echo Cancellation) for music playback or rendered audio responses,
noise suppression in diffuse fields, music and multi-talker environments,
beamforming and source localisation to focus on the active speaker,
automatic gain control to absorb proximity variations from the user.

DSP IP licensing is a recurring point: several actors (Cirrus Logic, Knowles, DSP Group, Synaptics-CES heritage) offer commercial AFE bricks pre-tuned for Alexa Built-in or Cast. The choice of brick conditions the SNR achieved and the test-plan pass rate.

Anechoic chamber and diffuse-field testing

Acoustic tests are conducted in an accredited laboratory equipped with an anechoic chamber able to simulate reverberation profiles representative of domestic environments. The instruments are supplied by specialists such as Audio Precision, Klippel, Listen Inc. The standard sequence:

wake-word detection at variable distances (typically up to several metres),
multi-talker with competing voice sources at different angles,
AEC during music playback at various levels,
noise suppression with diffuse noise injection,
false-accept rate with non-target audio content.

The exact target values are defined in the Amazon test plan and are subject to change. The order of magnitude of effort, when starting from a non-tuned AFE, is several iterations of physical testing, with hardware adjustments (mic placement, acoustic ports, internal damping) between sessions.

Wake-word and licensing

The wake-word engine is a critical brick. Two patterns:

embedded third-party engine, the most common case for non-Amazon products (Sensory TrulyHandsfree is the historical reference),
proprietary Amazon engine, available under licence for certain partners.

The wake-word activation rate, the latency between user utterance and AVS streaming, and the false-accept rate per hour are all metrics covered by the test plan. The engine and its tuning ship as a deliverable that travels through certification.

Cloud connection and credential management

The AVS cloud connection works on a specific stack.

Stack overview

HTTP/2 endpoints on the Alexa cloud, with multiplexing for low-latency audio,
OAuth 2.0 through Login With Amazon for user authentication on the device,
refresh tokens managed by the device, with rotation according to Amazon's rules.

A common bug is refresh-token rotation that breaks after a device reboot: the device stores a refresh token, the reboot loses the rotated state, the cloud refuses the next refresh request, the user is logged out without explanation. This pattern is caught in pre-cert review when it appears.

Encoding and upload

The audio uploaded to AVS is typically encoded in Opus at 16 kHz mono, sometimes in PCM depending on the configuration. The choice between Opus and PCM influences bandwidth and codec latency, both factors weighed by Amazon during functional testing.

Privacy and microphone indicator

The Amazon Privacy and Security Requirements for Alexa Built-in impose strict rules on user privacy.

physical mute button required: a hardware control that physically disconnects or visibly inhibits the microphone array,
LED indicator when the microphone is actively streaming audio to the cloud, with an always-on policy during streaming,
microphone state signalled at every interaction, no silent listening,
factory reset that fully clears stored credentials and personal data.

The indicator-LED rule is regularly underestimated: a device that emits the LED only at the start of streaming and not for its full duration is non-compliant. Pre-cert review verifies the LED is on for the entire streaming window.

Google Cast: Chromecast built-in and Assistant built-in

Google's accessory programme covers two main product tracks, plus an emerging Matter extension.

Chromecast built-in for audio

Audio-only products: smart speakers, connected sound bars, audio receivers. The integration goes through the Google Cast SDK on the receiver side, with a sender on phone, tablet or browser (the Cast app and the Chrome Cast extension). The architecture:

mDNS-SD (Multicast DNS Service Discovery) for discovering receivers on the local network,
DIAL (Discovery And Launch) to launch the receiver application,
Cast Application Framework (CAF) for session and message management between sender and receiver.

The product also typically integrates Google Assistant built-in when it has a microphone, with an acoustic test plan close in spirit to Amazon's.

Chromecast built-in for TVs

TVs and connected displays. Add to the audio track:

video rendering with H.264, H.265 and AV1 codec support depending on the product class,
Widevine DRM, in L1 for HD or UHD premium content, in L3 for lower resolutions,
HDR pipeline with Dolby Vision or HDR10 plus depending on commercial positioning,
Dolby Atmos pass-through for compatible audio outputs.

Hardware requirements (RAM, CPU, AES decoder support, Wi-Fi class) are defined by Google in published documents, with minimums that evolve by SoC generation. The order of magnitude is a Wi-Fi 5 minimum with Wi-Fi 6 preferred on recent generations, but the actual figures applicable to a given project must be verified with Google.

Google Assistant built-in

A certification layer separate from Cast, dedicated to voice integration. The functional and acoustic test plan resembles in spirit the Amazon plan, with its own metrics: wake-word detection rate ("Hey Google" or "OK Google"), far-field SNR, AEC during music playback, false-accept rate. A product can be Cast-only, Assistant-only or both, depending on the targeted scope.

Matter casting

The 2024 to 2026 horizon introduces Matter casting: casting roles (Casting Source and Casting Target) on the Matter fabric. The point is to converge with the Matter ecosystem on the discovery and session-management layer, while keeping the Cast SDK on the application and media layer. The exact integration depends on Matter version revisions and ecosystem actor decisions: at the time of writing, the Cast over Matter angle is being structured.

Track	Type	Functional cert	Acoustic cert	DRM
Chromecast built-in for audio	Audio receiver	Yes	Often paired with Assistant if mic present	Lower constraint
Chromecast built-in for TVs	Video receiver	Yes	If embedded mic	Widevine L1 for premium HD or UHD
Google Assistant built-in	Voice integration	Yes	Yes, full acoustic test plan	Not applicable
Matter casting	Casting extension over Matter	Emerging	Depends on product type	Depends on product type

Widevine and the DRM track for video

For Cast products that render premium video, Widevine integration is a critical brick.

Widevine L1 versus L3

Widevine L1 (hardware-backed): keys handled in a Trusted Execution Environment (TEE), content decoding running in a protected path. Required for HD and UHD premium content from Netflix, Disney+ and equivalents.
Widevine L3 (software): keys and decoding in software. Acceptable for lower resolutions or specific configurations, insufficient for premium HD or UHD content.

KeyBox and provisioning

L1 integration requires loading a KeyBox in the SoC at the factory, in a process supervised by the SoC vendor and Google. The KeyBox is unit-specific and not transferable. Forgetting the KeyBox at factory, or losing it during repair flows, is a recurring failure mode that results in a non-playable device.

Frame drops at handshake

A frequent pre-cert symptom is frame drops at the Widevine handshake when the stream starts or at key rotation. The cause is typically a timing problem in the protected pipeline or a slow key derivation that does not absorb in the audio-video synchronisation budget. Resolution involves measurements internal to the SoC and tuning of the protected decoder.

For high-end audio products, no equivalent DRM layer applies: the audio chain does not carry equivalent content protection.

Cross-programme considerations

Several constraints apply to both AVS and Cast.

Regulatory radio certification

AVS and Cast certifications do not replace radio regulatory certification. A connected speaker for the European market needs:

CE marking under RED (RED radio compliance overview),
Wi-Fi Alliance certification (Wi-Fi Alliance certification),
Bluetooth SIG qualification if Bluetooth is used (Bluetooth SIG qualification),
AVS or Cast certification under the Amazon or Google programme.

For the United States, FCC Part 15 takes the role of CE under RED. For Japan, MIC certification with the giteki marking, and so on for the other regulatory regions. The radio layers stack on top of the programme layers, not in substitution.

IoT cybersecurity

Privacy and security requirements imposed by Amazon and Google on connected accessories partly overlap with public referentials such as NIST IR 8425 and EN 303 645, notably on update mechanisms, default-password management and vulnerability handling. The convergence is not total, but well-organised cybersecurity work covers both the Amazon or Google ask and the public referentials.

Acoustic IP licensing

The choice of AFE brick (DSP, wake-word engine, beamforming) involves third-party licensing. The relevant actors at the time of writing include Cirrus Logic, Knowles, DSP Group, and Synaptics with the CES Conexant heritage. The licensing structure and the technology roadmap evolve, and a current verification with each vendor is the only solid source.

Matter as a unifying extension

Matter is gradually introducing itself as a unifying layer for the connected home, with extensions on the casting side for Google and on the smart-home side for both Amazon and Google. The Cast over Matter angle and Alexa Smart Home Matter support are tracked in parallel by both programmes. Matter does not replace the voice or media layer: it adds to it.

Common pitfalls

Pitfall	Consequence
AFE with insufficient microphone SNR	Wake-word detection rate below acoustic-test-plan threshold
AEC failing during music playback	Wake-word masked, dialogue blocked while music plays
OAuth refresh-token rotation broken after reboot	User logged out, cloud connection lost without explanation
Cast video frame drops at Widevine handshake	Stream rejected, premium content non-playable on the device
Mic mute LED not always on while streaming	Privacy rule breach, fails Alexa Privacy and Security review
Factory reset that removes Cast keys without re-provisioning	Device unusable as Cast receiver after reset
Confusing Alexa Built-in and ACK during scoping	Wrong SDK chosen, integration restarted at programme switch
Wake-word IP not licensed at industrialisation	Forced licence under time pressure or engine reselection
Underestimating Wi-Fi Alliance and Bluetooth SIG certifications on top	Programme certification obtained but product not radio-conformant
Forgetting the Widevine KeyBox at factory	L1 chain non-operational, premium HD or UHD non-playable

Step-by-step procedure for a smart-speaker product with Alexa Built-in

The typical sequence for a manufacturer integrating Alexa Built-in into a new product.

Open the Amazon Developer account and sign the Alexa Built-in agreement, with selection of the target track (Built-in, ACK, Built-in Lite).
Choose the SoC and the AFE brick, validate DSP IP licensing (DSP supplier, wake-word engine, beamforming).
Integrate the AVS Device SDK or AVS-IK, including OAuth via Login With Amazon and refresh-token management.
Acoustic and mechanical design: microphone array placement, internal damping, acoustic ports, mute button and LED.
Internal acoustic tuning with provisional measurements in an in-house or low-cost chamber, before passing to an accredited laboratory.
Pre-cert acoustic testing in an accredited laboratory (Audio Precision, Klippel, Listen Inc.), wake-word detection rate, AEC, multi-talker.
Pre-cert functional testing on the Alexa Functional Test Plan: dialogues, capabilities, response rendering.
Privacy and security review against the Amazon Privacy and Security Requirements: mute, LED, factory reset, data clearing.
Wi-Fi Alliance and Bluetooth SIG certifications in parallel, on the regulatory radio side.
Regional radio certification: FCC, CE under RED, MIC, KC and so on according to the target markets.
Submit the full dossier to Amazon, with acoustic and functional reports, privacy declarations, regulatory references.
Iterate on Amazon comments, conduct any complementary testing, get the Alexa Built-in pass.
Brand authorisation and access to commercial Alexa Built-in deliverables (logo, marketing, registry on the Amazon site).
Maintenance: track Alexa SDK updates, Amazon test-plan revisions, Wi-Fi Alliance and Bluetooth SIG roadmap.

For cross-cutting orders of magnitude per phase, see certification timeline.

Comparing AVS and Cast

Criterion	Alexa Built-in (Amazon)	Chromecast and Assistant built-in (Google)
Cloud entry point	AVS Device SDK or AVS-IK, HTTP/2 + OAuth via LWA	Cast SDK or Assistant SDK, mDNS-SD + DIAL + CAF
Wake-word	"Alexa", local engine + acoustic test plan	"Hey Google" / "OK Google", local engine + test plan
Acoustic test plan	Alexa Acoustic Test Plan, accredited laboratory	Equivalent Google acoustic test plan, accredited laboratory
Video DRM	Not applicable for typical audio products	Widevine L1 for premium HD or UHD on TV products
IoT lite track	ACK (Alexa Connect Kit) on Amazon-hosted module	No strict equivalent, indirect via Google Home and Matter
Privacy and indicator	Mute, always-on LED while streaming, factory reset	Equivalent rules under Google policy
Brand	Alexa Built-in logo and brand authorisation	Chromecast built-in / Google Assistant built-in logos

The two programmes share a strong family resemblance on the AFE, voice cloud and privacy parts. They diverge on the DRM for video angle (specific to Cast on TVs) and on the IoT-lite track (where Amazon has structured ACK with no exact symmetric on the Google side at the time of writing).