Model reference · open weights
MOSS-TTS-Local-Transformer is an open-weight audio or speech model from mudler, listed in the AxForge catalogue. AxForge can bring it up on EU-owned hardware for you on request — with the licence handled where one is required.
About
MOSS-TTS-Local-Transformer-v1.5 (GGUF) Brought to you by the LocalAI team. GGUF conversion of OpenMOSS-Team/MOSS-TTS-Local-Transformer-v1.5 for moss-tts.cpp: a C++17/ggml inference port of the OpenMOSS MOSS-TTS family that runs entirely on stock ggml with no Python, ONNX or torch at inference time. Architecture MossTTSLocal v1.5 is an RQ-Transformer text-to-speech model: - Global backbone: a 36-layer Qwen3 transformer over time (the text + speech prefix), hidden size 2560. - Local depth transformer: a 1-layer GPT-J block (LayerNorm, fused QKV, interleaved RoPE, SiLU MLP) over the codebooks within each frame. - Binary decision head: a direct 2-wide localtexthead on channel 0 that chooses continue (audio slot) vs stop (audio end) per frame. - 12 RVQ codebooks decoded through MOSS-Audio-Tokenizer-v2 at 48 kHz stereo (interleaved output). The non-matmul tensors (all lc. heads and embeddings, every norm and bias) are kept in f32 in every quant; only the Qwen3 global and GPT-J local attention/FFN matmuls are quantized, so the CPU gather/head paths stay exact. Verification This port is numerically verified against the reference PyTorch implementation: the language model matches to about 1e-5 with an exact greedy code sequence, and the MOSS-Audio-Tokenizer-v2 decode matches at about 115 dB SNR (bit-exact in f32). On CPU it runs roughly twice as fast per frame as the reference PyTorch at the same f32 precision. Files The non-matmul path (heads, embeddings, norms, biases) stays f32 in both quants. This repo also ships the codec (moss-audio-tokenizer-v2-f32) and tokenizer (moss-tokenizer-v15) GGUFs required to run the model. Usage With moss-tts.cpp built: The output is a 48 kHz stereo WAV. Swap --model for any of the quants above. Links - Inference engine: https://github.com/mudler/moss-tts.cpp - Base model: https://huggingface.co/OpenMOSS-Team/MOSS-TTS-Local-Transformer-v1.5 - Codec: https://huggingface.co/OpenMOSS-Team/MOSS-Audio-Tokenizer-v2 License Released under the Apache-2.0 license, following the upstream base model.
Summarised from the published model card. Read the full card on the HuggingFace links below.
Specifications
| Maker | mudler |
|---|---|
| Type | Audio & music |
| Variants | 1 |
| Based on | OpenMOSS-Team/MOSS-TTS-Local-Transformer-v1.5 |
| Released | 2026-07-12 |
| Popularity | 158k downloads / month |
| Likes | 1 |
| Licence | Open weights |
How it works
Variants
Open weights ship in several sizes and precisions. One page, all the variants — pick the one that fits your GPU. VRAM figures are estimates from model size.
| Variant | Params | Precision | VRAM | Fits 16 GB | Weights |
|---|---|---|---|---|---|
| MOSS-TTS-Local-Transformer-v1.5-GGUF | — | GGUF | — | — | Weights ↗ |
Using it via the API
Once AxForge deploys mudler-moss-tts-local-transformer for you, it answers on the OpenAI-compatible API — the same base URL and keys as every other model. (mudler-moss-tts-local-transformer below is illustrative; you get the exact model name on deployment.)
$ curl -sS https://api.axforge.ai/v1/audio/transcriptions \ -H "Authorization: Bearer $AXFORGE_API_KEY" \ -F model="mudler-moss-tts-local-transformer" -F file=@audio.mp3
Details
Tags
Licence
Open weights under apache-2.0 — commercial use is permitted. Deploy it on AxForge EU hardware on request. Read the licence ↗