SpiritBox Get Ultimate $199

SpiritBox Ultimate 1.12.12 / R7.12-rc1 · Windows Early Access

Guide-based conversations with your local AI.

Use Guides to set reusable conversation instructions and keep saved chats in one interface. Ultimate uses its bundled local engine by default, with LM Studio available as an optional backend. It is for technical Windows users with Docker Desktop and WSL2/Linux containers.

View offer · $199 USDCheck requirements

Consider Ultimate if you want to evaluate its Guide-based conversation interface and local setup. LM Studio already supports saved chats and reusable prompt presets, so those features alone are not a reason to buy another app. Review the bounded testing evidence before spending $199 USD.

One-time purchase. Models are separate. Existing checkout terms and purchase rights remain unchanged.

The download

What you receive.

The application, Windows setup scripts, bundled-engine quick start, optional LM Studio instructions and third-party notices. Models are separate downloads.

Current Windows package: SpiritBox_Ultimate_v1.12.12_R7.12-rc1_Windows.zip (82,739,363 bytes). See the package checksum and release evidence.

Start safely

New installation: extract into a new empty folder and follow QUICK_START.md. For the optional LM Studio backend, use LMSTUDIO_FIRST.md.

Existing installation: read the included UPDATING.md. Extract the complete new release into a separate new folder, keep Docker Desktop running, then run 4_Update_SpiritBox.bat from that new folder and provide the path to your existing installation. Review its plan before confirming. Do not start a second installation against the same container names or extract over the old folder. Back up configuration and user data, retain the updater backup and previous archive, and do not delete Docker volumes. Follow the Updating guide for complete backup and rollback instructions.

Broader workflows are separate

Optional services require separate setup and qualification. ComfyUI models, nodes and workflows and n8n are not supplied 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.

Testing evidence

What passed, and what remains unproven.

Current 1.12.12 / R7.12-rc1 checks

The exact 1.12.12 / R7.12-rc1 package passed isolated updater and browser-component checks. The updater used inert container commands and synthetic health responses; existing configuration, conversations and a custom Guide were preserved, and all 305 packaged app Python, JavaScript and template files matched manifest hashes. Guide provider selection, preview and saved changes, plus Telegram readiness states, were checked with controlled provider responses. F5 is optional and requires separate configuration; personal voices and reference files are not bundled. Video remains unqualified. A fresh full Docker installation and physical listening-quality sign-off were not completed. Seller-hosted buyer archive bytes have not been independently rehashed.

Historical 1.12.10 / R7.10-rc1 checks

The 1.12.10 / R7.10-rc1 archive passed repeatable-build and independent package verification (584 members, zero warnings or errors). Acceptance included 313 Python tests and 35 JavaScript assertions. Five original legacy images displayed again after reopening, with original conversation and image hashes unchanged; legacy tool provenance remains labelled unknown. Model-free media checks passed. This maintenance run does not qualify audible speech, heavy generation, clean-PC installation or other hardware. Seller-hosted archive bytes have not been independently rehashed.

Historical 1.12.9 / R7.9-rc1 checks

The archive passed repeatable-build and independent verification (580 members). Acceptance included 259 Python tests, 11 JavaScript policy assertions, 10 shipped replay-method assertions and authenticated synthetic save/reopen/replay/media checks. Voice identity and UI playback states were verified; audible listening and heavy media generation were not qualified. These checks do not establish clean-PC installation or other-hardware support. See the release evidence and boundaries.

Bounded real-model checks

The bundled Qwen3-14B route completed a checklist on R7.6 and preserved it through restart. Optional LM Studio GPT-OSS20B completed a factual summary on R7.7, then preserved the identical answer through reload, reopening and app restart.

Incomplete-answer handling

A captured empty model response was replayed offline on R7.7. The explicit incomplete-response notice persisted after reload and reopening. R7.7.1 fixes Windows launcher path quoting with six small edits across three batch files. All 56 lightweight launcher and preflight tests passed against the extracted hotfix. Prior R7.6/R7.7 runtime evidence is unchanged; a fresh installation with real Docker and a model was not rerun for this hotfix. No new voice generation was performed for the historical R7.7.1 hotfix. Read the full evidence boundary.

Before checkout

Check your existing setup.

Required setup

  • Windows with Docker Desktop using WSL2/Linux containers
  • Bundled spiritbox-llm engine, or an optional existing LM Studio model/server
  • Separately downloaded models suitable for your hardware

Qualification limits

Hardware needs depend on your model and configuration. This release does not establish a general minimum RAM or VRAM requirement.

Full-workstation features, other hardware, other models and long conversations remain unqualified.

Read the offer before you buy.

Review the full 1.12.12 / R7.12-rc1 requirements, testing limits and buyer policies on Gumroad.

View Ultimate · $199 USDBuyer policies

Also explore SpiritBox Pro or Basic.