What is Maya 2 Native best for?

Maya 2 Native is best for multilingual local speech generation.

Local TTS model

Maya 2 Native

Q: Can Maya 2 Native run locally?

Maya 2 Native is listed by LocalClaw as a local TTS option. Hardware fit depends on runtime, model size and backend support.

Catalogue summary: Streaming 2.50B text-to-speech model for 11 Indic languages and English, with 20 built-in voices and zero-shot cross-lingual voice cloning from 3-10 seconds of reference audio. Official Linux CUDA inference produces 24 kHz mono audio in 80 ms frames.

Repository editorial metadata; verify comparative claims in the linked upstream material.

GPU recommendedtext-to-speech generation12 languagesApache 2.0 weights/code; Mimi codec CC-BY-4.0; ECAPA Apache 2.0

Choose an app

Start with the recommended app, or choose another supported option.

Compare speech models

Open model filesRecommended verified starting point Open in UnslothLaunches Unsloth Desktop on this model Open official projectApplication source and guided setup Open on Hugging FaceFiles, licence and model card Check my machineOptional: verify the fit with your saved hardware

Desktop app links require the app to be installed. If nothing opens, LocalClaw will show app-download and model-file fallbacks.

Catalogue quality

Not rated

Catalogue speed

Not rated

Model size

2.50B unique parameters; approximately 5 GB BF16 weights, plus codec and speaker encoder

Voices

20 built-in voices; zero-shot and cross-lingual cloning with 3-10 seconds of reference audio and its exact transcript

Can Maya 2 Native run locally?

Maya 2 Native can generate speech locally for private voice workflows. Use the verified setup options on this page; no terminal command is required to choose the right path.

Apache 2.0 weights/code; Mimi codec CC-BY-4.0; ECAPA Apache 2.0 license. Still verify upstream usage notes before shipping.

Upstream source

multilingualvoice-cloningstreaming

Model details

Type

Local TTS model

Family

maya

Latency

not rated

Formats

pytorchsafetensors

Languages

as, bn, en, gu, hi, kn, ml, mr, or, pa, ta, te

Context

Official HF card and github.com/MayaResearch/maya2-native. BF16 backbone, FP16 depth decoder, 2048 KV tokens per seat: one seat uses 7.7 GB steady GPU memory and peaks at 9.2 GB during loading, plus about 0.5 GB CUDA context. Upstream README says 8 GB+, but its loading measurements exceed that; the model card recommends 12 GB or larger. Increasing max_seats raises memory usage. Cache model, Kyutai Mimi codec and SpeechBrain ECAPA encoder before offline inference. Assamese is cross-lingual for every built-in voice. Cloning quality varies and has no dedicated benchmark in this release. No local inference test performed.

Install locally

Check runtimeConfirm the backend supports pytorch, safetensors on your machine.

Open recommended setupUse the app and model links above. LocalClaw does not expose a terminal command.

Test locallyRun a short private audio prompt before moving into production workflows.

Good for

text-to-speech generation
GPU recommended local workflows
multilingual, voice-cloning, streaming

Watch before shipping

Validate pronunciation, latency and artifacts with your own voice samples.
Review the upstream license and acceptable-use notes.
Benchmark on your target CPU, Apple Silicon or GPU setup.

Related TTS and speech models

CompareBrowse all TTS models Local AIBrowse LLM models macOS appGet LocalClaw