Microsoft Azure Custom Neural Voice and Custom Avatar Limited Access

Properties

kind: provider-policy
target: voice-and-likeness
jurisdiction: Microsoft (industry)
mechanism: Custom neural voice and custom text-to-speech avatar are both Limited Access features usable only after a customer registers through Microsoft's intake form and is approved for a specific, pre-selected use case; before training a model the customer must obtain the talent's explicit written permission, share Microsoft's Disclosure for voice and avatar talent in advance, and submit a recorded acknowledgement statement from the talent, which Microsoft verifies against the training data using its own speaker-recognition biometric technology to confirm the voices (or, for avatar, the video) are from the same person.
status: active
owns: Distinct from every tracked provider-policy row by gating access at the application/registration level rather than the account level: a customer cannot attempt to build a custom voice or avatar model at all without Microsoft approving the specific use case in advance, and Microsoft independently biometrically verifies the talent's recorded consent statement against the training data itself rather than relying solely on the customer's self-report, unlike the ElevenLabs, Resemble, Descript, Cartesia, Hume, and Synthesia rows which all rely on the customer's own representation or an account-based verification step.
source: Microsoft Learn, Limited Access, Custom neural voice and custom text to speech avatar
notes: Page metadata shows a last-updated date of 2026-08-26, shortly before this lens's 2026-09-04 to 2026-09-19 window; no confirmed edit within the window itself was found, so this is included as an untracked provider policy rather than an in-window change. The parallel custom text-to-speech avatar clause on the same page requires an acknowledgement video verified against training video rather than audio. Customer is explicitly stated to be responsible for ensuring all necessary permissions are obtained from voice or avatar talent.

Provenance

Source: https://learn.microsoft.com/en-us/azure/ai-foundry/responsible-ai/speech-service/text-to-speech/limited-access

Properties

kind

provider-policy

target

voice-and-likeness

jurisdiction

Microsoft (industry)

mechanism

Custom neural voice and custom text-to-speech avatar are both Limited Access features usable only after a customer registers through Microsoft's intake form and is approved for a specific, pre-selected use case; before training a model the customer must obtain the talent's explicit written permission, share Microsoft's Disclosure for voice and avatar talent in advance, and submit a recorded acknowledgement statement from the talent, which Microsoft verifies against the training data using its own speaker-recognition biometric technology to confirm the voices (or, for avatar, the video) are from the same person.

status

active

owns

Distinct from every tracked provider-policy row by gating access at the application/registration level rather than the account level: a customer cannot attempt to build a custom voice or avatar model at all without Microsoft approving the specific use case in advance, and Microsoft independently biometrically verifies the talent's recorded consent statement against the training data itself rather than relying solely on the customer's self-report, unlike the ElevenLabs, Resemble, Descript, Cartesia, Hume, and Synthesia rows which all rely on the customer's own representation or an account-based verification step.

source

Microsoft Learn, Limited Access, Custom neural voice and custom text to speech avatar

notes

Page metadata shows a last-updated date of 2026-08-26, shortly before this lens's 2026-09-04 to 2026-09-19 window; no confirmed edit within the window itself was found, so this is included as an untracked provider policy rather than an in-window change. The parallel custom text-to-speech avatar clause on the same page requires an acknowledgement video verified against training video rather than audio. Customer is explicitly stated to be responsible for ensuring all necessary permissions are obtained from voice or avatar talent.