We've Released Voice Canvas as OSS: Easily Create Your Favorite Voice

Описание и применение

Voice Canvas – это открытый инструмент для преобразования голоса, работающий без GPU и позволяющий пользователям настраивать восемь параметров голоса через слайдеры, автоматически выбирая и смешивая до четырёх референсных голосов.

Задача

Users need a way to transform voice timbre while preserving speech content, tempo, and intonation without requiring a GPU or complex setup.

Результат

Voice Canvas enables users to create custom voices on CPU, with a 3‑minute demo video and 69 bundled anchors. Conversion times range from 2 to 3.5 minutes for 3–10 second clips.

Рабочий процесс

• Clone the Voice Canvas repo
• Clone the Seed‑VC repo
• Create a Python 3.10 virtual environment
• Install dependencies
• Run the app
• Select source audio (TTS, mic, or file)
• Adjust sliders or graph points
• Convert to generate takes
• Fine‑tune speed/pitch if needed
• Export WAV

Доказательства

Voice Canvas is an app that changes voice timbre while preserving the original audio's speech content, tempo, and intonation.

Источник

It runs without a GPU, though conversion on a CPU takes between one to several minutes per execution.

Источник

The package includes 45 speakers from Common Voice Japanese (which has clear terms of use) and 24 speakers from JVNV (Japanese Natural Emotional Speech Corpus)...

Источник