Check your machine before buying.
Ultimate R7.7.1 uses Windows and Docker Desktop with Linux containers and WSL2. The bundled spiritbox-llm engine is primary; LM Studio is optional. Models are separate and hardware requirements are model-dependent.
Start with Minimal and QUICK_START.md. LMSTUDIO_FIRST.md covers the optional external backend. Model choice, context length and other workloads determine hardware fit.
Pro: before you buy
SpiritBox Pro runs as two Docker services: the conversation app and Piper. On Windows 10/11 x64, use hardware virtualization and Docker Desktop with Linux containers and WSL2. On macOS, use an Apple Silicon Mac and Docker Desktop for Apple Silicon; WSL2 is Windows-only. Both downloads require LM Studio with a compatible local model and its server on port 1234. Model files are not included. First installation may download image and package dependencies.
Typed messages in; configurable Piper spoken replies out. Basic and Pro do not include microphone input, document memory, voice cloning or voice training.
Basic 1.3.2 downloads are available for Windows and macOS Apple Silicon. Both require LM Studio, a compatible local model, and its server on port 1234; no Docker is required. The compiled Windows build embeds Python and application libraries, so it does not require a separate Python installation. For Mac setup, follow the instructions included with the DMG. Intel Mac support is not offered.
Ultimate R7.7.1: setup requirements and limits
- Operating system
- Windows 10 or Windows 11. The current customer ZIP does not qualify macOS or Linux.
- Container runtime
- Docker Desktop with Compose, Linux containers and WSL2. Start with the documented Minimal profile.
- Chat backend and models
- The bundled spiritbox-llm engine is primary. LM Studio is optional. Model files are not included. A separate embedding model is not required for Minimal chat; Full semantic workflows require separate configuration and validation.
- GPU
- Hardware fit depends on the chosen model, context length and other running workloads. Configured-host checks do not establish a general minimum or qualify other GPU configurations.
- Memory and storage
- Model-dependent RAM and free disk space for Docker images, your chosen models and saved content. R7.7.1 does not establish a general RAM or VRAM minimum.
- Internet
- Initial downloads and optional integrations can require network access. Core local use does not require a SpiritBox cloud account.
Important limits
- The bundled Qwen3-14B checklist passed on R7.6. Optional LM Studio GPT-OSS20B summary and restart persistence passed on R7.7. An offline replay verified saved incomplete-answer notices. No new voice generation was performed.
- Full, other models, long conversations and other hardware qualification remain incomplete. Review the exact R7.7.1 release evidence.
- ComfyUI and n8n require separate setup; they are not provided as a configured, tested workflow bundle. Full retrieval/graph, VoiceLab training and live Telegram/n8n workflows require separate configuration and validation. n8n is user-managed.
- CPU-only, AMD GPU, Intel GPU, macOS, and Linux customer support are not included in this Windows release.
- Optional web search, Telegram, remote ComfyUI, and external APIs create network connections when enabled.