# Kenpath Labs

> Kenpath Labs is a voice AI lab building speech models in India, for the
> world. Its flagship is Svara TTS Turbo (model id svara-tts-turbo), a
> text-to-speech model in beta: hundreds of regional voices, 80+ languages in
> one model, code-switching mid-sentence, consent-gated voice cloning, inline
> expressive tags, input streaming, ~80 ms to first audio. The lab also ships
> Svara TTS v1, an open-source TTS foundation model (Apache 2.0, 850K+
> downloads on Hugging Face). Early access for founding teams is open.
> Contact: hello@kenpathlabs.com | https://kenpathlabs.com

Kenpath Labs originated from Kenpath and operates as its own lab in
Bengaluru, India. Positioning: people trust what speaks their language, and
buy from it — across markets, not only India. Voice AI that speaks the
customer's own language is the value; speed is how it holds a conversation.
The planned API platform serves Svara TTS Turbo only (v1 stays open source,
self-hosted).
Every audio demo on the site is real, unedited model output.

## Pages

- [Svara TTS Turbo](https://kenpathlabs.com/): the homepage is the Turbo
  page — voice sampler, code-switching reel, input-streaming race, voice
  cloning, replayed multilingual support calls, feature grid, platform plans.
- [Open source](https://kenpathlabs.com/open-source): Svara TTS v1 deep dive
  plus the lab's open-source tooling.
- [Early access](https://kenpathlabs.com/early-access): the founding-teams
  program — benefits, limited seats, and the interest form.

## Svara TTS Turbo (beta)

- 80+ languages in one model (28 Indic, 30 African, plus the world's
  majors); every voice speaks all of them, and voices stay identical across
  languages.
- Hundreds of regional voices; zero-shot voice cloning from a few seconds of
  reference audio, consent-gated by design.
- Code-switching mid-sentence (Hinglish and friends), inline expressive tags
  typed into the text ([sighs], [laughs harder], [clears throat]).
- Input streaming: the LLM's tokens flow straight into the model, so speech
  starts a few words into the reply. About 80 ms to first audio.
- A completely new architecture informed by v1's 850K+ downloads — not v1
  scaled up. Licensing not yet announced.

## Svara TTS v1 (open source)

- [Svara TTS v1](https://kenpathlabs.com/open-source): open-source TTS
  foundation model for 19 Indian languages in native scripts, with emotion
  conditioning. Apache 2.0 forever. 850K+ downloads on Hugging Face, peaked
  at #7 among TTS models, 13 community finetunes.
  - Model: https://huggingface.co/kenpath/svara-tts-v1
  - Demo space: https://huggingface.co/spaces/kenpath/svara-tts
  - GitHub: https://github.com/Kenpath/svara-tts-inference
  - Blog: https://huggingface.co/blog/kenpath/svara-tts-open-multilingual-speech-for-india

## Platform

API access to Svara TTS Turbo is at api.kenpathlabs.com
(console coming soon). The platform serves Svara TTS Turbo only — Svara TTS
v1 remains open source for self-hosting. Founding teams receive keys first,
with founding pricing locked in before public pricing exists.

## Optional

- [Get early access](https://kenpathlabs.com/early-access): limited seats
  for founding teams — early Platform and API access, a direct line to the research team,
  founding pricing, partnerships, and research collaboration.
- GitHub org: https://github.com/kenpath
- Hugging Face org: https://huggingface.co/kenpath
- LinkedIn: https://www.linkedin.com/company/kenpath-tech