I'm working with a non-profit that needs to store and playback a large number of audio samples of speech on a very small battery-powered device. Don't sweat all of that except the need for a very space-efficient, low-compute CODEC. We're looking at tradeoffs between processor complexity and FLASH capacity. Cost is extremely important. Our audio data set is about 145 minutes of mono voice, low-passed at 8kHz in Audacity before encoding. We'd like to hold to 10 to 12 bits of resolution to match the DAC we're using. We only need decode on the device, not encode. The 8kbps and 12kbps encoding and decoding tests I've done using the downloadable executables seem to produce good quality. The hard part is, for that kind of decode using Opus, how much processor do we need? It would help an awful lot if we could get by with an M0+ for cost reasons; lots of inexpensive ones with built-on DACs out there. Has anyone tried to get something like this to work on an M0+ processor, or otherwise looked at the performance requirements for NB-only low-bit rate voice only decode using only fixed-point? Thanks! Steve