Drama 3 Local Voice Studio is our independent interface for organizing local text-to-speech work. The published v0.1.1 release focuses on the work around a take: keeping a draft, reusing a voice reference, finding an earlier result and saving completed audio.
Start with the Forge download page. The release is source code that needs setup. It does not include a model, an engine binary or a ready-to-run Windows executable.
1. Check the prerequisites first
Read the release's Windows installation guide before downloading model files. The documented configuration uses Python 3.10+, Node.js 22.12+ during setup, compatible NVIDIA/CUDA/Visual C++ prerequisites and a trusted audio.cpp engine.
The UI code is Apache-2.0. Separately obtained model weights have their own terms; the UI license does not provide model rights or unrestricted commercial speech-generation rights. This project is not an official Fish Audio product.
2. Install the published source
Download the v0.1.1 source ZIP from the release page and extract it. Follow RELEASE.md for the setup command and model/engine paths. Use the release's checksum file to check the downloaded package. The standalone launcher does not require Pinokio.
Do not treat this as proof that every fresh Windows machine is supported: a complete clean CUDA/toolchain installation has not been validated. Existing-engine setup and synthetic generation were tested; the release documents the boundaries.
3. Build a repeatable voice workflow
Keep your text as a persistent draft, select a reusable voice reference you are authorized to use, and produce a short take before committing to a longer passage. The interface includes offline English writing checks and optional local Whisper dictation.
Use the searchable take history to compare versions. Change one part of the text at a time so you can judge whether the revision helped. Listen to the result before using it in a published video or sending it to someone else.
4. Save outside the studio without losing history
v0.1.1 adds Windows output-folder selection and Save as. The save path avoids overwriting existing files, and internal history retains its copy when an external destination fails.
After choosing a destination, make one short test take and verify that the expected file appears. Actual native-picker selection remains user-operated and pending confirmation in the published validation record. If saving fails, preserve the internal take and choose a reachable destination.
What is available today
The public release is v0.1.1. A newer local revision is not a published download yet. This guide describes the published version only; it does not claim that hosted Forge generation or token purchases are active.
Sources: v0.1.1 release notes, validation scope. Disclosure: Plops Forge publishes this interface and this guide.