Skip to content
← Models
Open weightsMeta/10 trustMetrics-checked
MMS TTS logo

MMS TTS

Text-to-speech for 1,100+ languages — Meta's Massively Multilingual Speech project.

Visit official page →

About

MMS (Massively Multilingual Speech, May 2023) is Meta's project covering speech synthesis (and ASR) in over 1,100 languages — an order of magnitude more than any other open TTS effort, including many low-resource and endangered languages with no commercial TTS coverage at all.

CC-BY-NC 4.0 — research/non-commercial only. Each language has its own checkpoint (e.g. facebook/mms-tts-eng for English). Quality varies enormously by language: high-resource languages are decent, while many low-resource checkpoints are intelligible but rough — the value is coverage, not polish. For accessibility and language-preservation work it is unmatched; for commercial products, the license rules it out.

Specs

Context window

N/A (text-to-speech model; per-language checkpoints)

Pricing

Free for research/non-commercial use only (CC-BY-NC 4.0). No commercial license available.

Free quota

Free — download per-language checkpoints for non-commercial use.

API access

Self-hosted (Hugging Face per-language checkpoints; Transformers support).

Weights

Open weights

Released

May 22, 2023

Benchmarks

Provider-claimed (Meta paper): coverage of 1,107 languages with intelligible synthesis — treat coverage claims as verified by the checkpoint count; quality claims are paper-reported. No standardized TTS quality benchmarks across the language set.

Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.

Use cases

ttsmultilingualaccessibilitylocalizationresearch

Trust score — 6/10

/10 trustMetrics-checked

Scored 6/10 (metrics tier). Unmatched language coverage (1,100+) makes it uniquely valuable for low-resource languages and preservation work. Capped by: (1) CC-BY-NC 4.0 — no commercial use; (2) wildly uneven quality across languages; (3) 2023-era with no updates. A public-good research artifact rather than a product building block.

How the trust score is built

Full methodology →
Adoption · stars/downloads, log-scaled vs peers
7.4/10
Maintenance · recency, releases, issue load
4.0/10
Openness · license + docs
4.9/10
Manual review · sanity read + red-flag check
6.0/10
  • · HF downloads 82,540 (peer-normalized)
  • · released 1240 days ago
  • · manual review: unmatched 1,100+ language coverage; quality wildly uneven by language; 2023-era; license: CC-BY-NC 4.0 — research/non-commercial only

Sources