# MOSS-TTS
MOSS-TTS is an open-source speech and sound generation model family from MOSI.AI and the OpenMOSS team, a research group under the Shanghai Innovation Institution (SII) working in close collaboration with [Fudan University](https://www.edgepedia.org/fudan-university) and MOSI Intelligence.[1](https://howaiworks.ai/models/moss-tts) • [2](https://github.com/OpenMOSS/MOSS-TTS/blob/main/README.md) The family covers far more than plain text-to-speech: stable long-form synthesis, multi-speaker spoken dialogue (MOSS-TTSD), voice and character design from text prompts alone (MOSS-VoiceGenerator), real-time streaming TTS for voice agents (MOSS-TTS-Realtime), and environmental sound effects (MOSS-SoundEffect).[2](https://github.com/OpenMOSS/MOSS-TTS/blob/main/README.md) All weights and code are released under Apache 2.0, though the training data is not.[1](https://howaiworks.ai/models/moss-tts)
The family is a foundation-model-era project: its first dialogue model appeared in 2025, the flagship grew to 8B parameters, and the record here runs through mid-2026. Vendor benchmark claims are reported as such below and, where they exist, set against independent readings of the same numbers.
| Fact | Value |
|---|---|
| Maker | OpenMOSS team (under SII, with Fudan University and MOSI.AI) [1](https://howaiworks.ai/models/moss-tts) |
| Model sizes | 1.3B to 8B parameters across nine checkpoints [2](https://github.com/OpenMOSS/MOSS-TTS/blob/main/README.md) |
| License | Apache 2.0, weights, and code; commercial use permitted [1](https://howaiworks.ai/models/moss-tts) • [5](https://www.open-moss.com/en/moss-ttsd/) |
| Realtime latency | 180 ms time-to-first-byte; 377 ms pipeline latency (vendor-measured) [2](https://github.com/OpenMOSS/MOSS-TTS/blob/main/README.md) |
| Downloads | 1,079,657 in 30 days as of 2026-07-08, sixth among TTS-tagged Hugging Face models [1](https://howaiworks.ai/models/moss-tts) |
| Training data | Not released; vendor describes a corpus of millions of hours [1](https://howaiworks.ai/models/moss-tts) • [3](https://d6108366.hf-mirror.com/papers/2603.18090) |
| Known limits | Long-horizon speaker drift in English; decoding sensitivity (default temperature 1.7) [3](https://d6108366.hf-mirror.com/papers/2603.18090) • [1](https://howaiworks.ai/models/moss-tts) |
## Release timeline and versions
The README lists nine released checkpoints, each published on [Hugging Face](https://www.edgepedia.org/hugging-face) and [ModelScope](https://www.edgepedia.org/modelscope): MOSS-TTS-v1.5 (8B, Delay architecture), MOSS-TTS 1.0 (8B), MOSS-TTS-Local-Transformer-v1.5 (4B), MOSS-TTS-Local-[Transformer](https://www.edgepedia.org/transformer) (1.7B), MOSS-TTSD-V1.0 (8B), MOSS-VoiceGenerator (1.7B), MOSS-SoundEffect (8B), MOSS-SoundEffect-v2.0 (1.3B DiT), and MOSS-TTS-Realtime (1.7B).[2](https://github.com/OpenMOSS/MOSS-TTS/blob/main/README.md)
The one precisely dated release is MOSS-TTSD v0.7 on 2025-11-01, which improved audio quality, voice cloning, and stability, added 32 kHz output, and extended single-pass generation length from 960 seconds to 1,700 seconds.[4](https://github.com/OpenMOSS/MOSS-TTSD) An independent check in July 2026 found no MOSS-TTS v2: the Hugging Face API does not resolve `OpenMOSS-Team/MOSS-TTS-v2` or `-v1.6`.[1](https://howaiworks.ai/models/moss-tts)
## Architecture and training as published
Everything in this section is vendor-reported, from the MOSS-TTS technical report.[3](https://d6108366.hf-mirror.com/papers/2603.18090) The family is built on MOSS-Audio-Tokenizer, a causal [Transformer](https://www.edgepedia.org/transformer-machine-learning-model) tokenizer that compresses 24 kHz audio to 12.5 frames per second using variable-bitrate residual vector quantization (RVQ) with unified semantic-acoustic representations.
Two generator architectures sit on top of the tokenizer. MOSS-TTS itself uses a delay-pattern design: a single Transformer backbone with multiple prediction heads and an RVQ-aware delay schedule, prioritizing structural simplicity, scalability, and a clean long-context operating point. MOSS-TTS-Local-Transformer adds a frame-local autoregressive module for stronger speaker preservation and shorter time to first audio. The report states the Local variant is generally stronger on zero-shot speaker similarity, while the delay-pattern backbone is more natural for duration control and ultra-long generation.[3](https://d6108366.hf-mirror.com/papers/2603.18090)
The training corpus spans millions of hours, built by converting open-domain recordings into single-speaker assets with cross-consistency gating on speaker consistency, language consistency, and transcript validity.[3](https://d6108366.hf-mirror.com/papers/2603.18090) The composition and provenance of that corpus are not published, and the data itself is not released.[1](https://howaiworks.ai/models/moss-tts)
Published capabilities include zero-shot voice cloning, token-level duration control, phoneme- and pinyin-level pronunciation control, Chinese-English code-switching, and stable generation up to hour-scale outputs.[3](https://d6108366.hf-mirror.com/papers/2603.18090)
## Benchmarks: vendor claims versus the same table read independently
OpenMOSS's README claims MOSS-TTS achieved state-of-the-art results on the open-source zero-shot TTS benchmark Seed-TTS-eval, "surpassing all open-source models and rivaling leading closed-source systems."[2](https://github.com/OpenMOSS/MOSS-TTS/blob/main/README.md) An independent analysis by HowAIWorks.ai, reading OpenMOSS's own published table, finds that claim unsupported: [Qwen3-TTS](https://www.edgepedia.org/qwen3-tts) 1.7B posts better English WER (1.50 versus MossTTSDelay's 1.84) and GLM-TTS-RL better Chinese CER (0.89 versus 1.37); ten of the seventeen open-source baselines in the table beat MOSS-TTS's Chinese CER.[1](https://howaiworks.ai/models/moss-tts)
Where MOSS does lead is speaker similarity, and on a smaller checkpoint. MossTTSLocal's 79.62 Chinese SIM is the highest figure anywhere in the table, edging closed-source Seed-TTS at 79.6, and its 73.28 English SIM leads all open models; the lead belongs to the 1.7B research checkpoint, not the 8B flagship.[1](https://howaiworks.ai/models/moss-tts)
Two further vendor claims need scoping. The claim that a MOSS model outperformed closed systems like Doubao and [Gemini 2](https://www.edgepedia.org/gemini-2).5-pro in subjective evaluations belongs to MOSS-TTSD, the dialogue model, not MOSS-TTS itself, and rests on OpenMOSS's own preference testing.[2](https://github.com/OpenMOSS/MOSS-TTS/blob/main/README.md) • [1](https://howaiworks.ai/models/moss-tts) And MOSS-TTS is absent from [Artificial Analysis](https://www.edgepedia.org/artificial-analysis)'s Speech Arena, so no independent blind human-preference Elo exists; as of the review, no independent replication of any published benchmark had been found.[1](https://howaiworks.ai/models/moss-tts)
## Realtime and agent story
The team positions its three generator lines as complementary: Delay emphasizes long-context stability, inference speed, and production readiness; Local emphasizes lightweight flexibility and streaming-oriented performance; Realtime models multi-turn context from prior text and user acoustics for voice agents.[2](https://github.com/OpenMOSS/MOSS-TTS/blob/main/README.md)
Vendor-measured latency figures are specific: MOSS-TTS-Realtime reaches a time-to-first-byte of 180 ms, and the sum of LLM-first-sentence time plus the Realtime TTFB is 377 ms.[2](https://github.com/OpenMOSS/MOSS-TTS/blob/main/README.md) These numbers come from OpenMOSS's own documentation; no independent latency measurement or published VRAM/hardware requirement exists for any family member.[1](https://howaiworks.ai/models/moss-tts)
## Licensing, availability and adoption
Weights and code for MOSS-TTS and MOSS-TTS-v1.5 are Apache 2.0 with no revenue cap, field-of-use restriction, or geographic carve-out; training data is not released.[1](https://howaiworks.ai/models/moss-tts) OpenMOSS separately states that MOSS-TTSD-V0 weights, inference code, and API interfaces are open source and support commercial use.[5](https://www.open-moss.com/en/moss-ttsd/) All checkpoints are published on Hugging Face and ModelScope.[2](https://github.com/OpenMOSS/MOSS-TTS/blob/main/README.md)
Adoption is measurable only in aggregate. Queried on 2026-07-08, `OpenMOSS-Team/MOSS-TTS` reported 1,079,657 downloads over the previous 30 days, sixth among text-to-speech-tagged models, behind Kokoro-82M (13,509,796), XTTS-v2 (9,828,098), [Chatterbox](https://www.edgepedia.org/chatterbox) (2,463,153), and two Qwen3-TTS checkpoints; Alibaba's five Qwen3-TTS repositories total 6,755,528 downloads, roughly 6.3 times MOSS-TTS's figure.[1](https://howaiworks.ai/models/moss-tts)
## Reception, controversies and limits
**The benchmark dispute** is the main controversy. OpenMOSS's "surpassing all open-source models" framing on Seed-TTS-eval is contradicted by the numbers in its own table on intelligibility, as detailed above; MOSS's defensible lead is speaker similarity on the 1.7B Local checkpoint.[2](https://github.com/OpenMOSS/MOSS-TTS/blob/main/README.md) • [1](https://howaiworks.ai/models/moss-tts) With no Speech Arena entry and no independent replication, the vendor's subjective "beat Gemini" claim for MOSS-TTSD remains unverified.[1](https://howaiworks.ai/models/moss-tts)
**Documentation and supply-chain concerns.** The model card warns that MOSS-TTS is sensitive to decoding hyperparameters, with an unusually high default temperature of 1.7. Both checkpoints require `trust_remote_code=True`, and no published hardware, training-data, or safety documentation exists: no voice-cloning consent mechanism and no watermarking. The independent review notes that an Apache 2.0 zero-shot cloning model accepting seconds-class reference audio is a straightforward impersonation tool.[1](https://howaiworks.ai/models/moss-tts) No misuse cases, lawsuits, or regulatory actions involving MOSS-TTS appear in the sources.
**Version drift.** OpenMOSS states that when the language field is omitted, v1.5 "may improve some languages and regress slightly on others compared with 1.0."[1](https://howaiworks.ai/models/moss-tts)
**Technical limits.** The technical report itself identifies long-horizon speaker drift, especially in English, as the dominant failure mode in ultra-long generation, and names stronger long-context speaker anchoring and broader low-resource language coverage as next directions.[3](https://d6108366.hf-mirror.com/papers/2603.18090) Emotion control quality and code-switching accuracy have not been independently measured.
## References
1. [MOSS-TTS — HowAIWorks.ai model review](https://howaiworks.ai/models/moss-tts)
2. [OpenMOSS/MOSS-TTS README (GitHub)](https://github.com/OpenMOSS/MOSS-TTS/blob/main/README.md)
3. [MOSS-TTS technical report (arXiv 2603.18090)](https://d6108366.hf-mirror.com/papers/2603.18090)
4. [OpenMOSS/MOSS-TTSD repository](https://github.com/OpenMOSS/MOSS-TTSD)
5. [Open MOSS — MOSS-TTSD product page](https://www.open-moss.com/en/moss-ttsd/)
---
*Topic: Encyclopedia › Technology and the built world › Computing and digital systems › Modern AI: foundation models, generative AI, and the AI industry › Model families and named models › Audio, music, and speech models*
*Initially written Sep 17, 2026 · Edited: Oct 11, 2026; Oct 15, 2026*
*Copyright 2026 EdgeChat AI, a subsidiary of Biostate AI.*
License: Edgepedia Community License 1.0, https://www.edgepedia.org/edgepedia/license