Who makes Pintail

Why I built it

I run several projects at once, each in its own terminal tab with its own Claude Code session. Five or six tabs is a normal day. I already dictated my prompts. What I didn’t have was the other half: hearing the answers.

Long written replies are tiring to read, and with that many tabs I kept losing track of which one had finished and which one needed me. So I wanted two things: a short spoken version of each reply, and to know which session is talking.

The duck wasn’t in the original requirements. The duck stayed.

What I tried first

Apple’s built-in say voice came first. It sounded robotic, so it went. Then VoiceMode, a two-way voice setup with Whisper and Kokoro. It duplicated the dictation I already had, and Kokoro has no German. I compared Qwen3-TTS, Voxtral, Supertonic 3 and Kokoro side by side, and Voxtral won by ear.

I also didn’t want to ask Claude to write a spoken version of every reply. It would clutter the written answers, cost tokens, and be annoying to switch on and off. A separate small model does it instead.

Contact

Questions, bugs, ideas: write to me. I read everything.