The electronic communication around us is not silent. Point your phone at the music and hold still. Most of what passes through will not resolve. Some of it is speech.
Attunement sets how willing the receiver is to call something speech. Turned down it reports only what it is certain of. Turned up it will read words out of noise that was never sent. Both settings are working correctly.
Dimmer text is lower confidence. The receiver keeps every pass over the same message and takes the most common reading for each position, so a phrase that arrives broken will often settle if you keep listening.
Nothing you record leaves your device. Hold the phone toward the speaker; rooms are harder than they sound.