I compared four options:

• No mask
• Kimtech N95 Duckbill disposable (my favorite!)
• 3M HF-802SD reusable respirator with speaking diaphragm.
• 3M 6200 previous-generation reusable respirator without speaking diaphragm.

I recorded myself on my MacBook with Audacity, reading the same paragraph in with each setup:

(wav).

(wav).

(wav).

(wav).

To me, the no-mask version is by far the clearest, but if I only heard the N95 version I probably wouldn't notice that anything was off. The two reusable respirators are the least intelligible by a good bit, and the speaking diaphragm helps a lot less than I'd hoped.

Here's a frequency plot (script, script):

Unfortunately, since I don't have a great sense of how important each frequency range is for understanding speech, I don't find this all that helpful. But maybe you do?

(One thing I do really like about the new version of 3M's reusable respirator, though, is the button for fit testing. You press the button and it closes off the filters, and if you're able to breath in then the mask is not sealed.)