Skip to content
Prompt Software
de

Prompt Software · Text to speech

Lectern: Text to speech with voice cloning for Mac

Turn text into natural speech on your Mac, in stock voices or your own.

Type or import a document, give each paragraph a voice and an emotion, and listen. Clone your own voice from a 15-second reading, or describe a new one. Save MP3, WAV or an audiobook with chapters. The speech model runs on your Mac's GPU, so your text and your voice stay on the Mac.

Languages
10
Model sizes
1.7B and 0.6B
Import formats
16
Interface languages
9
Add to cart
Lectern playing a four-paragraph document: the third paragraph, set to sound serious with a one-second pause, is highlighted while the fourth, read in a cloned voice named Thilo, is still being generated; the voice inspector shows the stock voice Ryan

What it does

  • Your own voice from one reading

    Read a short text that ends in a six-digit code. Lectern checks on the Mac that you read it, takes out background noise such as a computer fan, clones the voice and shows a similarity score. A German take is enough for the voice to speak English.

  • A voice and a mood per paragraph

    Excited, happy, confused, angry, serious, bored, sad, embarrassed, calm or whisper, or your own instruction such as "slowly, as if telling a secret". Speed, pitch, volume and pauses per voice or per paragraph.

  • Scripts with a cast

    Lines that start with a name and a colon become characters. Give each character a voice once, and the whole dialogue is read in those voices.

  • New voices from a description

    Describe age, tone and style, hear three candidates and keep one. A kept voice sounds the same in every paragraph. Or fine-tune a stock voice with an instruction such as "older and slower" and keep that.

  • Documents in, audiobooks out

    Import TXT, Markdown, PDF, Word, RTF, OpenDocument, EPUB, PowerPoint, Excel, HTML, the subtitle formats SRT, VTT, LRC and ASS, and Fountain screenplays, or open a whole folder as a project. Export MP3, WAV, M4A, FLAC, AIFF or an M4B audiobook with a chapter per chapter or per paragraph, plus SRT subtitles. Every file is marked as AI speech, in its metadata and with an inaudible watermark.

  • Listen while it renders

    Play a paragraph, a chapter or everything. Playback starts before the rest is generated, and batch tasks turn a stack of documents into audio files in one run.

Screenshots

  • The cloning page after a reading: verification passed with an 89 % match, a take report with 7 seconds of speech and 46 dB signal to noise, a voice name, gender and twelve avatars to choose from
    Cloning your own voice: read a short text with a six-digit code, Lectern checks the reading on the Mac and rates the recording. Add up to three takes, then start cloning.
  • The voice library on the Stock tab with nine voices, each with an avatar, language, gender and a play button; tabs for project, my voices, voice design, system voices, frequently used and favourites
    Choosing a voice: nine stock voices that speak ten languages, the macOS voices, and your own clones and designs, with search, language and gender filters.
  • A dialogue script with the characters Narrator, Emma and James, each line labelled with its character and avatar, and the voice cast panel assigning Uncle Fu, Vivian and Aiden
    Casting a script: lines that start with a name and a colon become characters. Pick a voice per character once, and the dialogue is read in those voices.
  • Voice design: a description of a warm, calm female voice in her forties, quick attribute buttons, a sample sentence and three generated candidate voices with Play and Keep this voice
    Designing a voice from words: describe age, tone and style, listen to three candidates and keep one. A kept voice sounds the same in every paragraph.

System requirements

  • macOS 15 or later
  • Mac with Apple silicon
  • 3 to 9 GB of disk space for the speech models, downloaded in the app
  • 16 GB of memory recommended for the 1.7B models

Third-party notices

Network access

Apart from the licence check, it connects only to these, and only for the feature named:

  • Hugging Face, to download the speech models when you ask for them

More from the same line

Cookies

This site needs no cookies to work. With your consent it loads Google Analytics to count visits.

Privacy policy