So as I'm pair-programming (vibe coding?) with Claude all day, usually with a browser open next to the terminal, every time I spotted a UI bug the routine was the same: screenshot → save → alt-tab to the terminal → drag the file in → type what's wrong → enter. Five steps across two windows, dozens of times a day, many times in quick succession. I am lazy, and so I built this tool that streamlines this for me.
SnippAI is a native macOS menu-bar app that lets you screenshot, drop directly into a voice box and send it into your terminal. I use this regularly, and I like it. Screenshot -> "fix the padding around this button" -> into your existing terminal session.
It uses FluidVoice to drop directly into a voice-to-text input as soon as you've taken the screenshot, although you can turn this off if you want to go straight text box.
If you're not familiar with Fluid, its a 100% local voice model that works really well and really fast. (I uninstalled wispr after using fluid). Everything stays local, there is no cloud. There is a one time model download, ~500mb, but its optional if you want to use the voice feature.
It's optimized for iterm2 and Terminal app. Claude desktop works too, but its a little janky, and I will probably be adding some other terminal adapters in the near future.
Hey HN,
So as I'm pair-programming (vibe coding?) with Claude all day, usually with a browser open next to the terminal, every time I spotted a UI bug the routine was the same: screenshot → save → alt-tab to the terminal → drag the file in → type what's wrong → enter. Five steps across two windows, dozens of times a day, many times in quick succession. I am lazy, and so I built this tool that streamlines this for me.
SnippAI is a native macOS menu-bar app that lets you screenshot, drop directly into a voice box and send it into your terminal. I use this regularly, and I like it. Screenshot -> "fix the padding around this button" -> into your existing terminal session.
It uses FluidVoice to drop directly into a voice-to-text input as soon as you've taken the screenshot, although you can turn this off if you want to go straight text box.
If you're not familiar with Fluid, its a 100% local voice model that works really well and really fast. (I uninstalled wispr after using fluid). Everything stays local, there is no cloud. There is a one time model download, ~500mb, but its optional if you want to use the voice feature.
It's optimized for iterm2 and Terminal app. Claude desktop works too, but its a little janky, and I will probably be adding some other terminal adapters in the near future.
Demo here:
https://x.com/brothaAkhee/status/2075601771020304700
Would love any feedback or suggestions!