# llms.txt for https://omnivoice.app
# This file provides guidance to LLM crawlers on what content and concepts
# are important on the Omni Voice website, to improve semantic SEO coverage.

# -------------------------------------
# ✅ Website Information
Website: https://omnivoice.app
Name: Omni Voice
Description: Omni Voice is a free, open‑source AI voice generator and zero‑shot text‑to‑speech model that supports 646 languages. It powers high‑quality multilingual TTS, zero‑shot Voice Cloning from a short audio sample, and text‑only Voice Design, with production‑grade speed and accuracy.

# Primary topics / entities:
# - "Omni Voice" AI voice generator
# - multilingual Text to Speech (TTS) in 646 languages
# - zero‑shot Voice Cloning from 3–25 seconds of audio
# - cross‑lingual Voice Cloning (clone once, speak any language)
# - text‑only Voice Design (describe a voice with text)
# - open‑source Apache 2.0 license, commercial use allowed
# - comparison vs ElevenLabs and PlayHT (WER, speaker similarity, speed)

# -------------------------------------
# ✅ Core Content Sections (based on actual page structure)
# These anchors and sections are safe and recommended for deep crawling.
Allow: /
Allow: /#hero
Allow: /#key-features
Allow: /#use-cases
Allow: /#comparisons
Allow: /#faqs

# -------------------------------------
# ✅ Important Legal / Policy Pages
# (URLs may be adjusted when legal pages are added.)
Allow: /terms
Allow: /privacy

# -------------------------------------
# ❌ Disallowed API or Internal Paths (implicitly via robots.txt)
# These paths are not user‑facing marketing content and should be ignored
# for content and reasoning.
Disallow: /api/
Disallow: /admin/
Disallow: /sign-in
Disallow: /sign-up
Disallow: /profile

# -------------------------------------
# 📌 Sitemap for better discovery
# The main sitemap lists all crawlable marketing and documentation URLs.
Sitemap: https://omnivoice.app/sitemap.xml

# -------------------------------------
# 🎯 Key Product Capabilities to Understand
# (Grounded in the actual Omni Voice homepage content)

# 1) Multilingual Text to Speech
# - 646‑language coverage from a single unified model
# - natural prosody and intonation across language families
# - pronunciation control via phoneme annotations and Pinyin
# - configurable speaking rate (approx. 0.5×–2.0× real‑time speed)

# 2) Zero‑Shot Voice Cloning
# - clone any voice from 3–25 seconds of reference audio
# - automatically extracts speaker timbre, accent, and rhythm
# - no training or fine‑tuning pipeline required
# - cross‑lingual: one cloned voice can speak any supported language
# - robust to noisy or imperfect reference recordings

# 3) Voice Design from Text
# - create a synthetic voice using only a text description
# - supports attributes like gender, age, pitch, accent, style, emotion
# - enables rapid experimentation with many voices without recording

# 4) Developer and Product Use Cases
# - audiobooks, podcasts, and long‑form content narration
# - game NPC dialogue and character voices
# - language learning and pronunciation tutors
# - customer support and virtual assistants
# - news, announcement, and broadcast‑style voices

# 5) Performance and Benchmarks
# - multilingual WER ~2.85% vs ElevenLabs ~10.95% (24‑language benchmark)
# - higher speaker similarity (SIM‑o 0.830 vs 0.655)
# - ~45× faster than real‑time on batch inference (RTF ≈ 0.022)
# - trained on ~581K hours of open‑source speech data

# 6) Licensing and Openness
# - released under Apache 2.0 license
# - free for personal and commercial use
# - open‑source implementation and research paper available on GitHub / arXiv

```
Canonical-Name: Omni Voice
Also-Known-As: OmniVoice, omnivoice.app
```