I printed this with a mixture of PLA and TPU - I used PLA to make the rigid front for holding the filter element, and the rest is soft for comfort on the face. It took MANY tries of settings to get printer set up to do TPU without stringing. When I was finally satisfied with the printer settings that worked well for the soft TPU filament, I used those settings for the whole mask. Started printing with PLA; paused the printer and changed the filament at 8 mm of z-axis travel. This got just past the dimples for the filter insert. Also, the sooner you transition between materials the more real estate you have for the materials' bonding. If you make the transition to TPU later, it has too little area for bonding with the PLA.
The insert for holding the filter element is also PLA.
Played around with retraction speeds and distance, Travel Speed, and very slow extrusion speeds - too little retraction causes TPU stringing, too much leads to jamming as extruder tries to "push a rope". I made test pieces with parallel 2mm thick walls to simulate the mask edges and guarantee travel between the sides to make sure there wasn't a stringing problem. Ended up with some smooth surfaces. Of course this was tuned for the Replicator 2 extruder, so I'd experiment with a test piece before committing to a large TPU piece. This piece took 8-10 hours - again, extruding slowly. (See Misc Notes)
Warhorse (PLA), NinjaTek (TPU) 3D Warhorse PLA filament, NinjaFlex TPU (Shore hardness 85A), All available on Amazon
PLA White, TPU Snow
All 30/50 mm/sec speeds were slowed to 20 mm/sec, 30% infill, 220 C, Travel Speed 175 MM/s, Retraction Speed 60 mm/s, Retraction Distance 3.3 mm, 2 shells, Roof & Floor Thickness .80 mm, Layer Height .20 mm. Co-worker (Thanks, Helen!) gave me the hint to cut TPU at an angle to facilitate loading. I was having trouble starting the feed. When switching to TPU, I applied some Canola or Peanut oil on the filament to lube the inside of the feeder tube. Soft filament wanted to bind in tube without it. And good luck !