PLOPS FORGE · VOICE STUDIO · OCTOBER 25, 2026

Voice selection and acting direction in Ming 3.0

Choose an existing voice and direct its delivery, or design a new voice from a description in Forge.

Open Ming voice studio

Voice identity and performance are different choices. In Forge’s Ming studio, selecting a voice does not hide the acting-direction field. Both remain on the same screen.

Choose a voice, then direct it

Use the “Choose a voice and direct its acting” approach. Select a built-in voice or a saved reference, then add an optional delivery description. “Warm, clear, and reassuring” is a useful starting point for an explainer. Combined reference conditioning and direction are experimental; audition the result.

Design a new voice

Choose “Design a new voice” when the description itself should determine the voice. Describe tone, texture, and delivery in plain language. The existing-voice picker is disabled in this mode because you are asking the model to design a voice.

Bring a reference

Use a mono PCM16 WAV recording lasting 1–30 seconds, up to 5 MB. Enter the exact words spoken and confirm permission to use the voice. Saved references are stored in this browser; they are not automatically synchronized between devices.

Use the same direction for a long take

Forge applies the selected voice, direction, and seed to every chunk before joining the recording. Chunk progress appears in the browser title. The studio shows the complete credit quote before you start.

Model source and documentation: inclusionAI/Ming-3.0-tts. The upstream source is MIT licensed.