Building Voice AI tools for serious consumers.
Now: @just_unmute — open-source dictation (free, BYOK, Sub 1 sec latency).
Next: Voice AI for serious athletes.
🚨 BREAKING: this guy is single-handedly killing WisprFlow.
Dictate, tell the AI to rewrite it any tone, style, or dialect you want. Done in under 1s.
WisprFlow: ~$12/mo, slower, closed source.
Unmute: free, open source. BYOK runs under $1/mo.
Pick a side.
Most voice AI builders treat always-on as a model problem. Get the model smart enough, and it handles everything.
Wrong.
After building multiple voice AI products, here's what I've learned:
A speech-to-speech model gives you probabilistic behavior. For some signals, that's fine. For others, you need determinism: if X happens, Y happens, always.
You cannot get that guarantee from a black box by making the black box better.
You get it by breaking the pipeline into stages you control.
STT → LLM → TTS isn't the workaround you throw away when models improve.
It's where control actually lives.
🚨 BREAKING: AI dictation tools are scamming users.
Their pitch: "AI-formatted output."
The truth: you're pasting into ChatGPT or Claude anyway. They don't need formatted prompts. Its just dumb to do so.
What you actually need: fast + accurate transcription. That's it.
WisprFlow: $12/mo for the wrong feature.
Unmute: free, open source. Nails latency + accuracy.
Pick a side.
My entire stack as a solo dev in 2026:
1⃣ Claude Code
2⃣ Unmute (Fast dictation)
3⃣ Tella (Video recorder)
Covers 90-95% of my work. Stopped buying specialists.
🧵
I AM single-handedly killing WisprFlow.
Dictate, tell the AI to rewrite it any tone, style, or dialect you want. Done in under 1s.
WisprFlow: ~$12/mo, slower, closed source.
Unmute: free, open source. BYOK runs under $1/mo.
Pick a side.
🚨BREAKING: this guy just embarrassed WisprFlow on long-form dictation.
Dictated for 3 minutes. Pasted in 1.2s.
A 10-second clip? Also 1.2s. Length doesn't matter.
WisprFlow chokes on long audio. And they charge $12/mo.
Unmute: free, open source, BYOK <$1/mo.
Pick a side. → https://t.co/khENwJxGAt
🚨BREAKING: this guy just embarrassed WisprFlow on long-form dictation.
Dictated for 3 minutes. Pasted in 1.2s.
A 10-second clip? Also 1.2s. Length doesn't matter.
WisprFlow chokes on long audio. And they charge $12/mo.
Unmute: free, open source, BYOK <$1/mo.
Pick a side. → https://t.co/khENwJxGAt