leosmith wrote:Since I was clearly wrong about there not being sufficient Swahili text with audio for me, I'm not as motivated to create the material. I might do it just to get some experience with creation though.
You might want to just give the speakers a topic at a time, let them figure out some ideas about what to say, and then record them as they speak naturally. Just as it is done in the speaking section of language exams sometimes. It'll be easier for both you and them to do this. Try asking Speakeasy about his experience trying to create some audio materials. He ended up running out of budget while trying to get it done, having underestimated how difficult it is for a typical (literate) native speaker to read something fluently and sound natural. People who can do that are less common than you'd think.
Recording spontaneous speech means that you'll end up recording false starts, filler words ("uh, y'know, uhhh"), and occasionally mistakes. Perhaps you can mitigate this by playing the recording to the natives and encouraging them to point out any obvious mistakes, so that then you can re-record those little bits. While editing the audio, you can remove most or all of the false starts and filler words as well.
By the way, I have found I'm actually unable to read something fluently without mistakes myself. I've made a few recordings of Spanish and Latin before, and what I do is to try to notice when I make mistakes as I'm recording, and simply re-read the sentence immediately. Then I edit out the bad sentence afterwards. This is always a lot better than trying to get the whole thing right in one go.