What it does
- Unlimited, offline, on the CPU. Once the models are downloaded, voicing never touches the network.
- Long jobs that survive anything. A whole book runs in the background and resumes after a quit, a crash or sleep. TXT, Markdown and EPUB in; chapters out.
- Word timing on every line, as SRT, VTT and JSON, so video tools and agents can place captions and cuts.
- Bad takes caught for you. An aligner flags skipped or invented words and re-rolls the line.
- Agents are first-class. Every operation is an MCP tool and a CLI command. Jobs an agent starts show up live in the window, where you fix one line and re-voice only that line.
- Pronunciation under your control:
[word](respelling),[word](/phonemes/)and[pause:600ms]right in the script, plus dictionaries. - Vietnamese done properly: 23 voices from the North, Centre and South, with English words mixed in.
- German, down to Grimm’s 1857 spelling (Studio): old spelling is read as today’s German and kept as written in the captions.
- Spanish, Latin American (Studio): four voices for learners and storytellers, with Quiroga’s 1918 spelling read as today’s Spanish and kept as written in the captions.
- French, from France (Studio): four voices for learners and storytellers, with numbers, dates, money and Roman numerals read out in French.
- Italian, from Italy (Studio): four voices for learners and storytellers, with numbers, dates, money and Roman numerals read out in Italian, and Collodi’s 1883 spelling read as written.
- Portuguese, Brazilian (Studio): four voices for learners and storytellers, with numbers, dates, money and Roman numerals read out in Portuguese, and Lobato’s 1920 spelling read as today’s Portuguese and kept as written in the captions.
- Machine listening (Studio, every voice language): a speech recogniser and the aligner listen to every line and mark the words worth a listen, so you can check a language you don’t speak.
- Exports as WAV (48 kHz mono 16-bit), FLAC, M4A and M4B audiobooks, each marked as AI-generated speech in its metadata.
Listen
The same paragraph in four free voices, voiced on a Mac's CPU.
af_heart · US Englishbf_emma · UK EnglishFree and Studio
Free
- 28 Kokoro English voices, US and UK
- 23 VieNeu Vietnamese voices
- Unlimited synthesis and long jobs
- Every export format and word timing
- Quality check, MCP and CLI
Studio
- Everything in Free
- 8 Chatterbox Turbo English voices, the most natural in PlainVoice
- Expression tags such as
[laugh]and[whispering] - Voice cloning with consent, Vietnamese and English
- 4 Chatterbox Multilingual German voices
- 4 Chatterbox Multilingual Spanish voices, Latin American
- 4 Chatterbox Multilingual French voices, from France
- 4 Chatterbox Multilingual Italian voices, from Italy
- 4 Chatterbox Multilingual Portuguese voices, Brazilian
- Machine listening for every voice language: English, Vietnamese, German, Spanish, French, Italian and Portuguese
- Perpetual: buy once, no subscription
Included with a PlainTake licence: one purchase, both apps.
For agents
Add PlainVoice to Claude Code as a plugin:
/plugin marketplace add plainvoice/plainvoice
/plugin install plainvoice
Or copy the setup for Claude Code or Codex from the app's Settings. Nine MCP tools cover
voices, models, lint, synthesis, jobs, line fixes and pronunciations, and every one has a
plainvoice CLI twin.