Back home

Luel Speech

sitemap 0.72
Visit source

Intro

Luel Speech is a text-to-speech product for underrepresented languages from voice-data company Luel. The public launch post: https://www.luel.ai/blog/luel-speech-is-now-available .

Context: commercial speech tech covers roughly one percent of the world's languages; when an Arabic dialect voice exists at all, it is often trained on the formal written standard, not spoken dialects. Ü-Tsang Tibetan and Dzongkha rarely appear on major platforms. Luel Speech targets that gap and is now publicly available.

Highlights:

  • Eight voices across four languages: Dolma & Tenzin (Ü-Tsang Tibetan, 48 kHz); Salma & Anir (Moroccan Darija, 24 kHz); Cleopatra & Mana (Egyptian Arabic / Masri, 24 kHz); Pema & Jigme (Dzongkha, 48 kHz);
  • Each voice built from its own native-speaker corpus, not a borrowed multilingual checkpoint;
  • Unknown voice slugs are refused rather than silently substituted;
  • Use via Luel Speech Studio (paste text, pick language/voice) or the API (TypeScript SDK @luel/speech);
  • Backed by Luel's contributor network (company states 900,000+ contributors across 96+ countries).

Explain

Problem

Most TTS stacks prioritize high-resource languages. Low-resource languages and spoken dialects are hard to ship without owning data pipelines. Luel Speech packages Tibetan, Darija, Masri, and Dzongkha as production voices callable from Studio or API.

Product shape

  • Studio: browser try-out with nothing to install;
  • API: one endpoint and one key; luel.speech.convert({ voice, text }) returns audio;
  • Voices are explicit slugs (e.g. bo-female in the docs example) so you get the voice you named.

Limits

  • Public language set is currently four languages / eight voices; out-of-list requests fail;
  • Sample rates differ by language (24 kHz or 48 kHz);
  • No invented latency or MOS scores here—use Luel's own docs for operational numbers.

Usage

Who it's for

  • Products needing TTS for Tibetan, Darija, Masri, or Dzongkha;
  • Voice AI, localization, education, and media teams integrating via API.

Getting started

  1. Read https://www.luel.ai/blog/luel-speech-is-now-available and try voices in Studio;
  2. npm install @luel/speech and set LUEL_API_KEY;
  3. Call speech.convert with a documented voice slug and your text;
  4. Stick to listed voices—do not assume silent fallback for unsupported languages.

Advertisement