In this video I show you exactly how I created a realistic voice clone (AI voice) inside Descript using plain text. I start by installing Descript on my computer (it works on both Windows and Mac), then I go to the Overdub section and click “Get Started.” The main requirement is simple: you need at least 10 minutes of your voice as a sample. I record live, but you can also import pre-recorded audio if you already have it. After recording, I transcribe the audio and I strongly recommend checking the transcript for mistakes. In my experience, the more accurate your transcript is, the better your cloned voice will sound—so I right-click words, use the “Correct” option, and make it as error-free as possible. Then I submit the recording + transcript as training data and record a short “voice ID” statement. This voice ID helps prevent other people from cloning your voice without permission and acts like your approval. Once Descript finishes training (usually less than a day), I create a new project, type any text, select my cloned voice from the speaker menu, and Descript generates speech that sounds very close to my real voice. Finally, I export the audio from Share → Export → Audio and save the final output.