The transcription service you actually want. Using latent semantic context (and a smorgasbord of tricks) to achieve SOTA transcription quality with speaker labels and timestamps for arbitrarily long audio — something that, to my knowledge, doesn't exist yet — plus optional readability enhancement.
I would be extremely happy to talk to someone about this project specifically. I genuinely think it's a good idea in need of a founder/founding team. I've just become too risk averse to make this a priority (seeing as I am probably going to lose health insurance in a few months and it currently isn't making me any money).
There is a public interface, but the website isn't hooked up to the current level of capabilities yet. Don't index too hard on it.
I just think they're whimsical and amusing.