@Christine Hanson Are you planning to use the Speaker recognition preview feature for authentication purposes in your application?
According to the documentation and guidance of this feature it is not intended to use the speaker recognition for these purposes.
The APIs are not intended to determine whether the audio is from a live person or an imitation/recording of an enrolled speaker.
With respect to the above questions, for a text dependent verification a specific passphrase should be used during enrollment and the same is used for verification. If you use the same recording then I think the verification will be successful.
For the second question, the same scenario should apply for text dependent verification. For text independent verification then atleast 20 seconds of total audio is required and a recording could be used successfully to verify the person.
Since the use-case for speaker verification is mostly based on improving remote meeting productivity by identifying users by voice profile, I think using the feature for authentication purpose is not appropriate.
If an answer is helpful, please click on or upvote which might help other community members reading this thread.