Text to Speech

Text-to-speech can create an editable audio item from new text or generate timed speech for existing captions.

Create a speech item

  1. Open the add menu on an audio track.

  2. Choose Text to Speech.

  3. Select the new item and enter its text in the inspector.

  4. Choose a model and configure the controls it provides.

  5. Generate the speech.

The available voices, reference inputs, emotion controls, and other settings depend on the selected model.

Generate speech for captions

Select caption clips and choose Generate Speech from the timeline context menu. You can also use the same action on a caption track to process all of its non-empty captions.

Choose a model that supports caption timing. Shrimply generates each caption at its existing duration and places the resulting audio on a track where it does not overlap existing clips.

Models and licenses

The current server supports Qwen3 TTS for built-in voices, voice cloning, and voice design, plus IndexTTS 2 and 2.5.

IndexTTS is provided under the separate bilibili Model Use License Agreement. Review that agreement before downloading or using the model; the Shrimply and compute-server licenses do not replace its terms.