But then I think this issue could be mitigated by integrating voice to text script. This would allow users to input their passphrase securely, without relying on keyboard input. It's not a perfect solution, but it could provide an additional layer of security for users who want to use passphrases.
No this wont help, this will actually only introduce even more security vulnerabilities and add multiple attack vectors:
- Same memory temporary storage vulnerabilities
- Likely introducing voice pattern analysis risks
- Network transmission interception
- All types of system audio recording malwares
- Audio processing services
TLDR: The most secure approach remains using dedicated hardware wallets or air-gapped devices, not adding more software layers that could be compromised.