Mellon

Speak, friend.

A voice that talks back like a person. Pick one and say what's on your mind.

Start with:

Best with headphones.

No custom model. Just timing.

The best voice companions train their own speech models. Sesame built CSM, and it’s remarkable. We wanted to see how close you can get without one. Mellon runs on off-the-shelf speech and language models, and we swap them freely. The human feel comes from everything in between.

  • Waits for you to finish.

    A pause to think doesn’t count as done.

  • Starts before you stop.

    It drafts a reply while you’re still finishing.

  • Stops when you cut in.

    Interrupt anytime. It goes quiet instantly.

  • Talks like a person.

    Short turns, real reactions, no lists, no lectures.

Replies in under a second.

Privacy

No audio stored. Calls last up to 5 minutes.