Wispr Flow: Dictation That Works EverywhereStop typing. Start speaking. 4x faster.
Promoted
iamhumans
iamhumans — teach a model to talk like a person
10 followers
iamhumans — teach a model to talk like a person
10 followers
It teaches a language model how to talk like a person. Not how to sound like one — that part's easy, and it's what most failures already do. This works on the shape underneath: when to be short, when to sit with something, when to push back, when the right reply is just "oh".
honestly this is a really interesting angle, like the focus on the rhythm and timing of conversation rather than just vocabulary feels pretty underserved. one thing i'd love to see is a way to test it against messy real scenarios, basically transcripts where the model has to navigate someone being vague or changing topics mid-sentence, since that's where most bots still fall apart for me.
Would love to see a side-by-side playground where I can compare the same prompt with and without iamhumans applied, so it's easier to actually feel the shift in tone rather than just trusting it.
Tried it on a few stubborn prompts and the pacing finally feels right — short replies when I was venting, a soft pushback when I was wrong, and a quiet "oh" that actually landed.
One thing I'd love to see is a small preview pane showing the raw prompt next to the reshaped reply, so I can see exactly what got softened, shortened, or pushed back on. Would make the underlying logic much easier to trust and learn from over time.
The framing of "shape underneath" really lands. Most tools stop at tone and miss timing, restraint, the small verbal nudges that make conversation feel real.
the framing here is genuinely sharp, going after conversational shape instead of surface polish is what makes this feel different from the usual "make AI sound human" pitches