Show a bug
Stop describing the bug. Show it.
Some things are quicker to point at than to write down. Press a hotkey, drag a box around the problem, and talk through what is wrong. You get an .mp4 and a plain-text transcript of your narration — so a coding agent reads what you said instead of decoding frames.
Drag a box, not the whole screen
Record the part that matters. Smaller files, and none of the rest of your desktop goes anywhere. Esc cancels and writes nothing at all.
Pause, then carry on
Pause stops the picture and the microphone together, so the gap is cut out of the finished clip rather than sitting in it. Your voice stays lined up with what it is describing.
The words come with it
Your narration is transcribed on your own machine and saved beside the video. Reading a few sentences beats watching two minutes of footage — for you, and for whatever you hand it to.
Straight into your project
Point it at a server and each clip is sent there as it finishes, using the SSH keys you already have — PipeVoice never stores one. Land it in the folder your agent is working in and it is simply there. Or keep everything local; that is the default.
Sounds like you meant it
Narration is captured at 48 kHz and levelled before it is saved, so a quiet mic or a loud one both come out clear. No more clips where the picture is fine and the voice is a whisper.
Buttons first, transcript after
Stop recording and the clip is yours straight away: play it, open the folder, send it, copy the path. The transcript is written in the background and follows a moment later. Nothing waits on it.
Files are named with the time in front, so a shared folder sorts by when things happened. Roughly 10–15 frames a second at 1080p: right for showing a bug, wrong for anything that needs to look smooth. More on using it with a coding agent →