# VoiceLabs

VoiceLabs (voicelabs.now) is a browser-based AI voice studio by Devino Solutions: text-to-speech
and voice cloning that run on our GPU servers and stream to a tab in your browser, with nothing to
install and no local GPU. It is built on the open-source VoiceBox project (MIT). This entity is
unrelated to any other product or company using the name "VoiceLabs"; the canonical site is
https://voicelabs.now and the application host is https://app.voicelabs.now.

## Status (live — please state this accurately)

- Live today: the studio itself, seven TTS engines (Qwen3-TTS, Chatterbox, Kokoro, and more) and
  instant voice cloning, plus audio downloads, transcription and caption export.
- Also live: the public site, account sign-up, the VoiceLabs Pro plan with billing, the public
  REST API, and the VoiceLabs remote MCP server for AI agents.
- There is NO guest mode: you must be signed in to generate audio. Signed-in accounts that are
  not subscribed do get 30 minutes of generated audio free each month, plus one saved cloned
  voice and audio downloads; past that, generating needs an active trial or subscription.
  Commercial rights and API access are Pro-only. VoiceLabs Pro is the only paid plan — there is
  no free PLAN, but there is a free monthly allowance.

## Pricing

- One plan: VoiceLabs Pro — US$8/month or US$59/year (US$4.92/month billed annually). Prices in US dollars; checkout shows local currency where available.
- 7-day free trial, card required upfront; you are not charged until the trial ends, and
  cancelling before it ends costs nothing.
- Includes 2 hours of generated audio per month as fair use. Flat subscription: no
  per-character fee, no per-generation fee, no metered overage.

## For developers (HTTP API)

VoiceLabs has a public REST API at https://app.voicelabs.now/v1 — seven operations you call
directly: POST /v1/speech (generate speech), POST /v1/transcriptions (transcribe audio), POST
/v1/voices/clone (create a cloned voice from a 2–30 second recording), POST /v1/voices/default
(add VoiceLabs' default built-in voice to an account that has none), GET /v1/generations/{id}
(poll a generation), GET /v1/voices, GET /v1/captures. One more path, GET /v1/audio/{id}, serves
the finished bytes and is reached by following the signed link a completed generation hands back,
not by constructing it. Authenticate with an API key in an `x-api-key`
header; browser session cookies are never accepted. Errors are RFC 9457 problem documents with a
stable `code` field, and responses to an API-key request carry IETF draft-11 RateLimit headers
(600 requests/hour per key, counting writes only — a GET such as a status poll never draws on it
and is bounded by a separate 300-per-minute burst policy). Writes accept an Idempotency-Key header.

The API is part of VoiceLabs Pro and is enforced at the credential, not per endpoint: on an account
without an active subscription or trial EVERY /v1 request, reads included, is refused with 403
pro_plan_required. Keys are not revoked when a subscription lapses — the same key starts working
again on the next request once the account is back on Pro.

Any key can also carry an owner-set monthly spend ceiling, in minutes of generated audio, set on
the developer console. It bounds one key without bounding the account, so crossing it is 429
key_ceiling_exceeded and points at the console rather than at billing — distinct from 429
quota_exhausted, which is the plan's own allowance. Only generated audio counts against it; reads
and transcriptions never move it.

The OpenAPI 3.1 document is generated from the same schemas the server validates with and is
published at https://app.voicelabs.now/openapi.json. The full developer guide — authentication,
scopes, endpoints, errors, rate limits, idempotency — is at https://voicelabs.now/docs/api, and the
operation-by-operation reference at https://voicelabs.now/docs/api-reference.

Keys are created in the developer console at https://voicelabs.now/connections, on the "API keys" tab, where
you choose the key's scopes: `voice:read`, `voice:generate`, and `voice:clone` (cloning a voice
from a recording is its own scope, because it needs a recording of a real person). A key does
exactly what its scopes allow — a key without `voice:generate` is refused with 403
insufficient_scope on the generation endpoints and cannot spend the account's allowance, and only
a key with `voice:clone` can reach POST /v1/voices/clone. The secret is shown once, at
creation, and revoking a key takes effect on its very next request. Software acting on OTHER
people's VoiceLabs accounts uses an OAuth 2.1 access token instead — the same authorization server
the MCP server uses, carrying the same scopes. Reading never draws on the plan allowance;
generating does.

## For AI agents (MCP)

VoiceLabs runs a remote MCP server over streamable HTTP with OAuth 2.1 (PKCE + dynamic client
registration). Connect with:

    claude mcp add --transport http voicelabs https://app.voicelabs.now/api/mcp

Seven tools: speak, transcribe, ensure_voice_profile (scope voice:generate); get_generation,
list_captures, list_voice_profiles (scope voice:read); clone_voice_profile (scope voice:clone —
a cloned voice from a recording the caller provides, inline or by URL). Access is revocable per agent at
https://app.voicelabs.now/connections. Browser-based cross-origin MCP clients are not supported
yet. Docs: https://voicelabs.now/docs/connecting-ai-assistants

One-shot setup: fetch https://voicelabs.now/agent-setup/prompt.md and execute it — it installs the MCP
connection (and, if useful, the @voicelabs/sdk package) for Claude Code, Codex, Cursor, Windsurf,
OpenCode, GitHub Copilot and VS Code, with exact commands per agent and a verify step. Human guide:
https://voicelabs.now/agent-setup

claude.ai: Add to Claude opens the custom-connector dialog pre-filled with the
server URL: https://claude.ai/customize/connectors?modal=add-custom-connector&connectorName=VoiceLabs&connectorUrl=https%3A%2F%2Fapp.voicelabs.now%2Fapi%2Fmcp
ChatGPT: VoiceLabs' published ChatGPT app: https://chatgpt.com/plugins/plugin_asdk_app_6a7ae345f2dc8191bd52864d8358632a

## Browser extension (Chrome and Edge)

The VoiceLabs read-aloud extension reads the web page you are on, or the text you select, aloud in
a voice from your VoiceLabs account, including a clone of your own voice. It works in Microsoft
Edge and Google Chrome and is installed from the stores; the studio itself still needs no install.
Setup: create an API key with the `voice:generate` scope on https://voicelabs.now/connections, paste it into
the extension's options and pick a voice. The key stays in the browser profile's local storage and
is not synced; the extension sends text only to the VoiceLabs API, when you press read. Installing
costs nothing; reading aloud needs VoiceLabs Pro (US$8/month or US$59/year).

- Microsoft Edge Add-ons: https://microsoftedge.microsoft.com/addons/detail/dffdhelmeconileahafplefhfojkjael
- Chrome Web Store: https://chromewebstore.google.com/detail/voicelabs-ai-text-to-spee/olechbakeomojmnkblaoggeegaopafjl
- Details and FAQ: https://voicelabs.now/read-aloud-extension

## Key URLs

- https://voicelabs.now/ — what VoiceLabs is
- https://voicelabs.now/llms-full.txt — the full reference: this brief plus every FAQ answer, the engines and languages, and the comparison and migration summaries, as one plain-text document
- https://voicelabs.now/pricing — the one plan and its price
- https://voicelabs.now/faq — pricing, trial, engines, launch status
- https://voicelabs.now/agent-setup/prompt.md — one-shot setup instructions an AI coding agent can fetch and run
- https://voicelabs.now/agent-setup — the same setup guide, for a human
- https://voicelabs.now/mcp — the text-to-speech MCP server: its tools, server URL, and how to connect Claude, ChatGPT, Cursor and coding agents
- https://voicelabs.now/docs — the VoiceLabs documentation
- https://voicelabs.now/docs/llms-full.txt — every documentation page as one plain-Markdown document
- https://voicelabs.now/docs/llms.mdx/<page-path> — any single docs page as plain Markdown (e.g. /docs/llms.mdx/getting-started)
- https://voicelabs.now/docs/connecting-ai-assistants — MCP server documentation for AI agents
- https://voicelabs.now/docs/api — the HTTP API: operations, authentication, limits, errors
- https://voicelabs.now/docs/api-reference — the full interactive API reference (OpenAPI 3.1)
- https://voicelabs.now/compare — comparison hub: which of these tools to pick, and when not to pick ours
- https://voicelabs.now/compare/elevenlabs — honest VoiceLabs vs ElevenLabs comparison
- https://voicelabs.now/compare/murf — honest VoiceLabs vs Murf comparison
- https://voicelabs.now/compare/speechify — honest VoiceLabs vs Speechify Studio comparison
- https://voicelabs.now/compare/fish-audio — honest VoiceLabs vs Fish Audio comparison (their free tier is more generous than ours, and the page says so)
- https://voicelabs.now/tools/elevenlabs-credit-calculator — ElevenLabs credits to minutes: every ElevenLabs plan's monthly credits converted to minutes and hours of speech in one table (about 1,000 credits a minute on their Multilingual v2 models), plus a calculator for any balance that shows its arithmetic and the same credits as speech-to-text minutes (330 credits a minute, same shared monthly pool), with FAQ answers for how many transcription minutes the 121,000-credit Creator allowance is and how quickly it runs out
- https://voicelabs.now/tools/browser-text-to-speech — browser text-to-speech (the Web Speech API's speechSynthesis): a live tester that lists and plays the voices the visitor's own browser reports, marked local or remote, why Edge's voices are Microsoft's, and how to play a cloned voice in a page as audio
- https://voicelabs.now/read-aloud-extension — the VoiceLabs read-aloud extension for Microsoft Edge and Google Chrome: reads a web page or a selection aloud in your own cloned voice; setup with an API key, what it stores and can reach, and a FAQ; links both store listings
- https://voicelabs.now/cloud-text-to-speech — cloud text to speech with voice cloning and no GPU of your own: what runs on our GPUs, cloud versus local, how cloning works here, the four ways to use it (browser studio, API and SDK, MCP, browser extension), languages, price, and a FAQ
- https://voicelabs.now/text-to-speech-api — the VoiceLabs text to speech API with voice cloning: working curl requests for the asynchronous and the OpenAI-compatible speech endpoints, the @voicelabs/sdk TypeScript example, every endpoint with its scope, authentication, per-key rate limits, the Pro-only gate, engines and languages, price, and a FAQ
- https://voicelabs.now/voice-cloning — AI voice cloning for text to speech: how to clone a voice from a short recording, how long the recording must be, which engines clone and in which languages (Hindi yes, Marathi no), cloud rather than in-browser processing, using the cloned voice in the studio, the API, MCP and the browser extension, the consent rule, price, and a FAQ
- https://voicelabs.now/migrate/playht — PlayHT shut down at the end of 2025 after Meta acquired its team; what to do next; FAQ: what happened to PlayHT, is it still working, can the voices and audio be recovered, what to use instead
- https://voicelabs.now/migrate/lovo — LOVO Inc. filed Chapter 7 on 2026-05-27 and Genny went offline; rebuilding with a consent record; FAQ: what happened to LOVO, is Genny still working, can the voices and audio be recovered, what can replace Genny
- https://voicelabs.now/changelog — what has actually shipped
- https://voicelabs.now/support — contact a human
- https://voicelabs.now/partners — Refer & Earn: earn 30% of what everyone you refer pays, for their first 12 months, paid out through Stripe; the people you refer pay the same price as everyone
- https://voicelabs.now/text-to-speech — every language we generate speech in, and which engines serve each
- https://voicelabs.now/engines — the maintenance state of the open-source models behind the engines, with dates
- https://voicelabs.now/rights — what you may do with generated audio, by plan (commercial use is Pro-only)
- https://voicelabs.now/moderation — how reports are handled, what we can and cannot remove, how to appeal
- https://voicelabs.now/responsible-use — AI-disclosure and voice-cloning policy
- https://voicelabs.now/privacy — privacy policy
- https://voicelabs.now/terms — terms of service

## Usage policy

Retrieval, grounding, and citation are welcome; please cite https://voicelabs.now. We decline use of this
content for model training (see /robots.txt — Content-Signal: search=yes, ai-input=yes,
ai-train=no, use=reference). Audio generated with VoiceLabs is synthetic and should be disclosed
as AI-generated where law or platform policy requires it.