AMD support is engine- and backend-specific.
VoiceStudio runs multiple voice engines. Its official performance guide says Windows PyTorch GPU acceleration is NVIDIA/CUDA-only: AMD graphics can therefore fall back to CPU for PyTorch engines. The same guide notes that the audio.cpp path can use a Radeon through Vulkan. Check the selected engine rather than assuming the whole app uses the GPU.
What this means on Windows
If an engine uses a PyTorch build that expects NVIDIA CUDA, installing that engine on an AMD-powered Windows PC does not make it use the Radeon automatically; CPU fallback may be the available route. For an engine using audio.cpp, the official guide describes a Vulkan route for Radeon GPUs. This is not a blanket promise that every model, driver, or VoiceStudio release will accelerate on every AMD card.
How to check whether your AMD GPU is active
- Open VoiceStudio’s Settings → Performance and inspect the reported compute device.
- Check the selected engine’s routing or GPU status; treat “CPU fallback” and “unknown” as not GPU-verified.
- Run a short representative generation and watch GPU utilization and memory in the operating system’s performance monitor.
- Compare elapsed time with a CPU run if the app lets you select the device. Keep the engine, model, text, and settings the same.
- Only download larger models or schedule long jobs after the result confirms a working GPU path.
AMD GPU compatibility is not the same as a universal minimum spec
GPU support can change by engine and app release. Driver versions, available VRAM, model size, and backend builds also matter. The safest answer is the live device report for your exact engine and machine—not the GPU brand alone. If the selected path is CPU-only, expect slower inference and test a short sample before relying on it for production.
Check the rest of your system
RAM, workload, and engine choice matter alongside the graphics card.
Read VoiceStudio system requirements