http://Handy.computer is a free and open source dictation app very similar to Wispr flow. Not quite as good IMO but more than good enough! I use handy on my personal life and Wispr flow for my professional life
I tried a bunch, including Handy, and unfortunately here I think there's a big difference between getting it 98% right (Handy) and getting it 99.99% right (Wispr, these are my numbers, not real).
The difference is that for Handy I had to revise everything, and for Wispr I don't.
Also Handy was really bad at handling language switches (needed for me as I move from English to Spanish and French consistently), so it's a no go. Wispr does that amazingly.
Clearly you are pushing the boundaries and getting great use, but you are also casually dropping $200/month *each* for 3 providers (maybe more?)
For people wanting to experiment with different models without committing such a big spend, openrouter maybe a valuable option. Also, I'd highly recommend moving your memory system to your own hard drive (call it "second brain") because then it becomes accessible to all your models. This works great when you use Claude Code (instead of Chat), Antigravity (instead of Gemini) or Codex (instead of ChatGPT).
Happy to riff a bit more (I am a practitioner -- look me up)
Yes, so not casually at all. This is why I was sharing the numbers: I'm doing it so that people know what they can get from each. My guess is that if I had to spend $200, it would be in either Anthropic or ChatGPT, and I'd try to get all tools to work there. If you're using any of these for work, the spend of at least 2 is widely justified.
Thanks for the tips on OpenRouter and local memory!
Great article with lots of useful tips.
http://Handy.computer is a free and open source dictation app very similar to Wispr flow. Not quite as good IMO but more than good enough! I use handy on my personal life and Wispr flow for my professional life
Thanks, good to know!
I tried a bunch, including Handy, and unfortunately here I think there's a big difference between getting it 98% right (Handy) and getting it 99.99% right (Wispr, these are my numbers, not real).
The difference is that for Handy I had to revise everything, and for Wispr I don't.
Also Handy was really bad at handling language switches (needed for me as I move from English to Spanish and French consistently), so it's a no go. Wispr does that amazingly.
Agreed, I am also a polyglot Wispr fan.
Clearly you are pushing the boundaries and getting great use, but you are also casually dropping $200/month *each* for 3 providers (maybe more?)
For people wanting to experiment with different models without committing such a big spend, openrouter maybe a valuable option. Also, I'd highly recommend moving your memory system to your own hard drive (call it "second brain") because then it becomes accessible to all your models. This works great when you use Claude Code (instead of Chat), Antigravity (instead of Gemini) or Codex (instead of ChatGPT).
Happy to riff a bit more (I am a practitioner -- look me up)
Yes, so not casually at all. This is why I was sharing the numbers: I'm doing it so that people know what they can get from each. My guess is that if I had to spend $200, it would be in either Anthropic or ChatGPT, and I'd try to get all tools to work there. If you're using any of these for work, the spend of at least 2 is widely justified.
Thanks for the tips on OpenRouter and local memory!