Describe a voice
Specify general vocal traits and make new original voice candidates rather than copying someone else's identity.
Kira World / Legacy & Developer Tools
From the research lab · Final legacy download
Kira World is moving toward more advanced, continuous synthetic-person architectures. Voice Designer 0.1.6 is the only standalone release planned for the older voice generator: the current free unsigned preview is that edition, not a preview of some later final. Kira World is developing a separate virtual Voice Box for its future synthetic people.
Final legacy version: 0.1.6 · Free unsigned developer preview available. No further Voice Designer versions are planned; full GUI qualification is incomplete.

Final legacy edition · Free Windows preview available
Version 0.1.6 is the final planned version of Kira Labs Voice Designer. The existing unsigned 0.1.6 installer and source archive are the only standalone legacy-generator edition Kira Labs plans to offer creators and programmers. Kira World's separate virtual Voice Box is a future research direction, not yet a qualified replacement. Evelyn’s approved original greeting is publicly playable, and Video Studio may continue using a preserved version of the older generator. No 0.1.7 or other new Voice Designer versions are planned. The available 0.1.6 EXE remains an unsigned experimental preview with untested graphical controls and no verified clean-PC installation.
Concept design · Proposed interfaceA creative workflow, not just a voice model
Four functions envisioned for the free edition, with a straightforward local interface and clearly documented exports.
Specify general vocal traits and make new original voice candidates rather than copying someone else's identity.
Enter text, audition generated samples and choose the sound you want to keep. Final user-interface behavior still requires testing.
Keep voice design profiles and export WAV audio for authorized creative, video, game or accessibility projects.
The planned tool will run separately from Evelyn and Kira World, with no companion memory or private resident data bundled.
What is verified: KiraWorld's Qwen3-TTS Voice Forge documents bounded local synthesis feasibility and audition artifacts. Its unsigned 0.1.6 preview installer has now passed backend generation and reinstall on one Windows test computer, but the new GUI and general release remain unqualified. The separate Local Voice Studio also has unfinished real-backend isolation requirements.
October 25, 2026 · Recorded backend tests · unsigned preview available
A preview executable and corrected source archive have been uploaded to the public Kira GitHub repository. Installation, model provisioning, uninstall, reinstall and real installed-backend voice generation were reported as successful on one existing Windows 11 test computer. The executable is unsigned, and the new graphical interface has not been tested interactively.
All eleven local VoiceDesign files passed integrity checks during setup and reinstall.
The installed command-line backend generated two distinct 2.72-second and 3.92-second synthetic samples. The earlier WAV survived reinstall unchanged.
Generate, Play and Save WAV buttons, clean Windows, unsigned-installer reputation, alternate hardware and missing-runtime setup remain unqualified.
Review boundary: An unsigned developer preview is not a certified release. The test host used 32 GB RAM and a 16 GB NVIDIA RTX 5060 Ti; these are observed specifications, not proven minimums. CPU-only speech is not implemented. The binary/source ZIP have not been independently safety-audited here, and no new private voice recordings have been placed on this page.
Final Voice Designer version · free unsigned Windows developer preview
These are the real installer and corrected-source files uploaded by Kira Labs, not pictures or placeholders. 0.1.6 is the final planned version; there will be no successor releases of this older generator. The installer has passed backend generation and reinstall on one Windows 11 test computer, but the new graphical interface and clean-machine compatibility remain untested.
Important before downloading: This unsigned EXE has not received an independent executable security audit or general-public release certification. Do not disable Windows security or override an untrusted-program warning to install it. Initial installation downloads approximately 4.52 GB of voice-model files plus other dependencies. A compatible NVIDIA CUDA GPU is required for the currently tested backend; CPU-only speech is not implemented. The tested host had 32 GB RAM and a 16 GB RTX 5060 Ti, but these are not certified minimum specifications. Generate, Play and Save WAV buttons in the new GUI still require testing.
Expected SHA-256 checksums, copied from the owner's supplied October 25 manifest. These have not been independently recalculated against the GitHub binary here, and unsigned checksums do not authenticate a publisher.
90892ff56e82247e5d3ae3a65631853332e9e4cf0757a633ce46772e5f9a320d0370d183099297c79622e7dbb49ac1341a26fff15b978bafab0ad3aec02498fcDirect downloads are pinned to the October 25 repository upload commit, so later source changes cannot silently change these exact links. See the installation and troubleshooting guide and third-party license notices before testing.
Release preparation · October 24 draft
The current Windows experiment has provisional GPU and memory checks; recommended hardware remains a planning target. Review installation symptoms, WAV and Chatterbox issues without mistaking draft specifications for certified minimums.
Early installer preview tested on one host / broader qualification pending
The current unsigned preview installer has provisioned its own Python, dependencies and local VoiceDesign model on one existing Windows computer. More hardware, clean-Windows installations, missing-runtime behavior and final release safeguards remain to be tested.
Check the computer — Windows version, architecture, available disk space, RAM and supported GPU/CPU route.
Download and verify — the selected model, exact software dependencies, permitted voice assets and license notices.
Install and test — first speech generation, WAV export, clean restart and an actual disconnected-computer test where offline use is promised.

For creators and programmers

Alongside a future installer, Kira Labs intends to provide a getting-started guide, voice creation examples, WAV export instructions, hardware requirements and troubleshooting information. Developers should be able to understand what is reproducible and what remains experimental.
A separate guide will explain how Kira Labs used voice-generation assets with Chatterbox, distinguishing tested steps from experimental or unsupported integration. Compatibility will not be advertised unless the matching steps have been reproduced.
Any public edition will contain only rights-cleared models, code, licenses, and authorized generic or original examples. Evelyn's private voice, Synthetic Robert's recordings, protected Kira World voice profiles and other private reference packs are not part of a public release.
Archive policy and limitations · Final legacy version
Approve and preserve her actual voice and demonstrate that her installed runtime can reproduce it.
Package only the public voice designer and its required files; keep private character voices and personal records out.
Review third-party rights and notices, dependency licensing, download integrity, local security and consent safeguards.
Verify installation, output quality, restart behavior, supported hardware, offline behavior, uninstalling and accurate help instructions.
Final-version decision, October 25, 2026: Kira Labs will preserve and offer Voice Designer 0.1.6 as the final and only planned standalone legacy-generator release. Its already available unsigned preview is that edition: no separate polished final build, 0.1.7 or new feature-upgrade cycle is planned. The downloaded preview is already available; the checklist above explains why it must not be represented as a fully tested, signed or widely compatible release. The planned virtual Voice Box is a separate Kira World research direction intended to take the place of the generator; it is not yet proven working and is not another Voice Designer release. Private recordings and current supported uses of older voice components remain independent of the free download.