Manual Index

Home · Manual index

TTS Server User Manual

This is the user manual for Portable TTS Server V2, a linux distro for windows users. The complete release runs from a local Windows folder and uses its own WSL2 disk. Kokoro and all app runtimes are included; other engines install into that disk when selected.

A GitHub source checkout has no runtime image. Download the complete release or follow the build instructions.

You can also read the online GitHub wiki, with the same illustrated guides and a navigation sidebar.

Read this manual when you want to install engines, speak text, clone a voice, generate SRT, edit audio, or drive the same surfaces from tts.cmd or the authenticated HTTP API.

Start here: the illustrated quickstart follows a real Kokoro narration from installation to a saved edit.

How to use these pages

Start with identity and requirements if you have never launched the app. Then follow the start/stop page, then the GUI tab that matches what you want to click. The CLI and HTTP API pages cover the same product from the command line and from another local app. The engine catalog is the A-to-Z of every registered model id in MODEL_SETUP. The dedicated A to Z knowledge page indexes every usage topic alphabetically.

Developer-oriented files such as CLI.md, API_REFERENCE.md, PORTABILITY.md, app_intents.md, and skills/use-tts-server/ remain in the app root for agents and maintainers. This manual/ directory is the user book.

Page index

Page	What it covers
Illustrated quickstart	A first narration: install, load, generate, edit, save, and download, with real app screenshots
Identity and requirements	What this product is, the linux distro for windows users, and every requirement needed to use it
Start, stop, and portability	Start-TTSServer.cmd, -VerifyOnly, first registration of ext4.vhdx, Shutdown, copying the tree
Storage, layout, and ports	E:\tts_server vs /opt/tts_server, jobs, voices, secrets, registry, ports 8300/9300
GUI overview	Window chrome, six tabs, gateway badge, toasts, how the WebView GUI authenticates
Setup tab	Hugging Face token, Install / Install All / Remove, install statuses
Server tab	Gateway panel, spawn workers, device, precision, Kill / Kill All
Voices tab	Upload, play, transcribe, rename, delete reference audio in voices/
Testing tab	Worker, Voice, reference audio, parameters, Generate, history
Editor tab	Library, waveform, transport, range ops, effects rack, Render
Log tab and Shutdown	Live log filters, Clear, the Shutdown button
Windows CLI (tts.cmd)	Every verb, discovery, exit codes, generation, jobs, audio editor
Authenticated HTTP API	Token, discovery registry, generate / jobs / voices / setup / workers / SRT / editor / maintenance
Engines overview	How install, load, speak, clone, and unload work across engines
Engines catalog	Every MODEL_SETUP id: bark, chatterbox, dia, f5, fish, higgs, kokoro, qwen, vibevoice, whisper, xtts, speecht5, parler, outetts, vits, edge, voxtral, voxcpm2, csm, orpheus
Jobs and projects	Job folders, recovery, chunk edit/rerun, save_path, deliver_to, projects
SRT and Whisper	Word-timed captions, source-guided vs ASR-only, Whisper sizes
Audio editor usage	Effect names, parameters, CLI tts.cmd audio, API /api/audio/*
Diagnostics and maintenance	diagnose, disk, env, gc-jobs, clear-cache, stale workers
Troubleshooting	Distro/path failures, 401 token, GPU reserve, missing Python, Edge offline
A to Z knowledge page	Alphabetical full-usage index of every topic in this manual

First session in one paragraph

Confirm Windows + WSL2, extract the complete release to a local folder, and double-click TTSServer.exe. When the GUI appears, open the Setup tab, paste a Hugging Face token if you will install gated models, check that bundled Kokoro is ready, switch to the Server tab, spawn a worker, switch to Testing, type text, click Generate. Finished audio lands under E:\tts_server\output\jobs\ unless you asked for a named project under E:\tts_server\projects_output\.

Conventions used in this book

• GUI labels are written as they appear on the button or tab: Setup, Server, Voices, Testing, Editor, Log, Shutdown, Install, Spawn Worker, Generate, Render.
• Engine ids are the lowercase keys the API, CLI, and MODEL_SETUP share: kokoro, xtts, edge, and so on. Display names such as "Kokoro 82M" are what the Setup cards show.
• Example paths use E:\tts_server and the legacy name linbox-TTS_Server; substitute your actual folder and the distro returned by Start-TTSServer.ps1 -VerifyOnly. New copies use a name derived from their path.
• Windows paths use backslashes (E:\tts_server\voices). Linux paths inside the distro use forward slashes (/opt/tts_server, /mnt/e/tts_server).
• Ports: the FastAPI gateway listens on 8300; the Windows WebView talks through the bridge on 9300. The CLI defaults to the bridge. The registry file lists both.
• Commands shown as tts.cmd ... are meant to be run from E:\tts_server in cmd.exe or PowerShell. tts.py is the same program if you invoke Python directly.

What this manual does not replace

It does not teach general Linux administration, covers moving the stopped folder in the portability page, and does not dump OpenAPI as a substitute for usage. Live OpenAPI remains at http://127.0.0.1:9300/docs and http://127.0.0.1:9300/openapi.json while the server is running.