OmniVoice is a massively multilingual zero-shot text-to-speech (TTS) model supporting over 600 languages. Built on a novel diffusion language model-style architecture, it delivers high-quality speech with superior inference speed, supporting voice cloning and voice design. - 600+ Languages Supported: The broadest language coverage among zero-shot TTS models. To get started, install the omnivoice library: You can use OmniVoice for zero-shot voice cloning as follows: For more generation modes (e.g., voice design), functions (e.g., non-verbal symbols, pronunciation correction) and comprehensive usage instructions, see our GitHub Repository. You can directly discuss on GitHub Issues. You can…
GEnerative, Prosody-aware, Autoregressive text-to-speech model for Realtime Dialogue Gepard is a text-to-speech model built for real-time conversation.
Parameters556M
Context262,144
Weights1.1 GB
Licenseapache-2.0
AccessOpen weights
Monthly Downloads69.2k
Runs On
What it takes to serve gepard-1.0 (556M parameters): the memory its weights need at each precision, and the cheapest way to rent enough data-center GPUs to hold them.
| Precision | Weights | Memory needed | Cheapest setup | Per hour | Also fits |
|---|---|---|---|---|---|
| 16-bit | 1.1 GB | 1.3 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
| 8-bit | 0.6 GB | 0.7 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
| 4-bit | 0.3 GB | 0.3 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
Memory is the weights at that precision plus 20% for the runtime and a short context; a long context needs more. Prices are the lowest on-demand hourly rates in the SAVRN Index, read Sep 18, 2026.
Model Card
By NineNineSix, published under apache-2.0, revision 440c3a27797e.
[](https://discord.gg/NzP3rjB4SB) [](https://opensource.org/licenses/Apache-2.0) [](http://arxiv.org/abs/2609.04222) [](https://www.nineninesix.ai/) [](https://huggingface.co/spaces/nineninesix/gepard)
Gepard
[!IMPORTANT] 2026-08-06 — updated
stop_headweights. Every other parameter is unchanged. This fixes premature stopping: multi-sentence inputs are now carried through to the end instead of cutting off at the first sentence boundary, and the effective duration ceiling is lifted. Short-phrase behaviour is unaffected.
GEnerative, Prosody-aware, Autoregressive text-to-speech model for Realtime Dialogue
Gepard is a text-to-speech model built for real-time conversation. It starts speaking the moment text begins arriving, generating audio piece by piece instead of waiting for a full sentence — so it feels like a live voice, not a recording. It's a single language model that learned text and speech together, so the output carries natural rhythm and timing rather than the flat, stitched tone of older pipelines.
The name evokes "Gepard"(/geh-PART/), German for cheetah — a nod to the model's low-latency, high-throughput streaming.
Configuration
- Architecture
- Qwen3_5ForCausalLM
- Context length (tokens)
- 262,144
- Layers
- 14
- Hidden size
- 1,024
- Feed-forward size
- 3,584
- Attention heads
- 8
- Key/value heads
- 2
- Head dimension
- 256
- Vocabulary size
- 248,320
- Model type
- qwen3_5_text
Identity and Version
- Repository
- nineninesix/gepard-1.0
- Publisher
- NineNineSix
- Task
- Text to speech
- Modality
- Audio
- Library
- transformers
- Parameters
- 556M parameters
- Languages
- en, es, pt, nl
- Revision
- 440c3a27797e98673c8e56f0569a86c09ea5de34
- First published
- 2026-06-22
- Last updated
- 2026-09-10
Files and Weights
11 files, 1.1 GB in total. The weights are 2 files totalling 1.1 GB in pt, safetensors.
Weights2 files · 1.1 GB
Configuration2 files · 5.8 KB
Tokenizer2 files · 20.0 MB
Documentation2 files · 25.4 KB
Other2 files · 2.5 MB
Repository1 file · 1.7 KB
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| after_DPO_stop_head.pt | Weights | 4.0 KB | fff769159daa |
| model.safetensors | Weights | 1.1 GB | 119651934205 |
| config.json | Configuration | 1.4 KB | — |
| gepard_config.json | Configuration | 4.5 KB | — |
| LICENSE | Documentation | 11.3 KB | — |
| README.md | Documentation | 14.1 KB | — |
| gepard_techreport.pdf | Other | 1.9 MB | 0f8dc388ac3b |
| logo.png | Other | 584.9 KB | 9a91aa495cb4 |
| .gitattributes | Repository | 1.7 KB | — |
| tokenizer.json | Tokenizer | 20.0 MB | 87a7830d63fc |
| tokenizer_config.json | Tokenizer | 1.1 KB | — |
License and Download
- License
- apache-2.0
- Access
- Open weights, no gate
- Download size
- 1.1 GB
Download from NineNineSix
Released by NineNineSix through its official repository on Hugging Face. Read the license.
Built From
- Derived from nineninesix/qwen3_5-full-attn-only-14
- Described by arXiv:2004.11362
- Described by arXiv:2110.13900
- Described by arXiv:2207.12598
- Described by arXiv:2212.04356
- Described by arXiv:2301.12597
- Described by arXiv:2305.18290
- Described by arXiv:2307.08691
- Described by arXiv:2309.06180
- Described by arXiv:2309.15505
- Described by arXiv:2405.14734
- Described by arXiv:2406.17957
- Described by arXiv:2501.15907
- Described by arXiv:2505.19462
- Described by arXiv:2506.09827
- Described by arXiv:2508.05835
- Described by arXiv:2609.04222
- Trained on (disclosed) laion/Emolia
Memory Requirements
| Precision | Weights in memory |
|---|---|
| As published | 1.1 GB |
| 16-bit | 1.1 GB |
| 8-bit | 0.6 GB |
| 4-bit | 0.3 GB |
Weights only, from the published parameter count; the key-value cache and runtime add to this.
Questions About gepard-1.0
How much GPU memory does gepard-1.0 need?
About 1.3 GB at 16-bit and 0.3 GB at 4-bit: the weights (556M parameters) plus a working margin. A long context needs more.
What is the cheapest GPU to run gepard-1.0 on?
At 16-bit, 1x MI300X from $1.85 an hour; at 4-bit, 1x MI300X from $1.85 an hour, at the lowest on-demand prices the SAVRN Index lists.
Can I use gepard-1.0 commercially?
Yes. gepard-1.0 is released under Apache License 2.0. The Apache License 2.0 is a permissive open-source license. It permits commercial use, modification and redistribution. It requires keeping the license and copyright notices and any NOTICE file, stating significant changes, and it includes an express patent grant from contributors.
What is gepard-1.0's context length?
262,144 tokens, from the maximum position embeddings in its published configuration.
Similar Models
VieNeu-TTS-v2 is the next generation of Vietnamese TTS, designed for Natural Communication, Podcasts, and Bilingual (En-Vi) Code-switching. This project features the flagship VieNeu-TTS-v2 architecture: Tác giả: Phạm Nguyễn Ngọc Bảo Training high-quality TTS models requires significant GPU resources. If you find this model useful, please consider supporting the development: Install the SDK to integrate VieNeu-TTS-0.3B into your research or applications: Deploy VieNeu-TTS as a high-performance API Server (powered by LMDeploy) with a single command. Start the Server with a Public Tunnel (No port forwarding needed): Once the server is running, you can connect from anywhere (Colab, Web Apps…
Qwen3-TTS is a series of advanced multilingual, controllable, robust, and streaming text-to-speech models developed by the Qwen team. This specific checkpoint is the 0.6B CustomVoice variant, based on the 12Hz tokenizer. It supports 9 premium timbres and allows for fine-grained style control over target voices via natural language instructions across 10 major languages. To use Qwen3-TTS, you can install the qwen-tts package: For Qwen3-TTS-12Hz-0.6B-CustomVoice, the following speakers are supported. We recommend using each speaker’s native language for the best results: If you find Qwen3-TTS useful for your research, please consider citing
Qwen3-TTS is a family of advanced multilingual, controllable, robust, and streaming text-to-speech models. Trained on over 5 million hours of speech data spanning 10 languages, Qwen3-TTS supports state-of-the-art 3-second voice cloning and description-based control. This specific checkpoint is the 0.6B Base model, which is capable of rapid voice cloning from a user-provided audio input. To clone a voice and synthesize new content using the Base model, you can use the following code snippet: Qwen3-TTS covers 10 major languages (Chinese, English, Japanese, Korean, German, French, Russian, Portuguese, Spanish, and Italian) as well as multiple dialectal voice profiles to meet global application…
Indic Parler-TTS is a multilingual Indic extension of Parler-TTS Mini. It is a fine-tuned version of Indic Parler-TTS Pretrained, trained on a 1,806 hours multilingual Indic and English dataset. Indic Parler-TTS Mini can officially speak in 20 Indic languages, making it comprehensive for regional language technologies, and in English. The 21 languages supported are: Assamese, Bengali, Bodo, Dogri, English, Gujarati, Hindi, Kannada, Konkani, Maithili, Malayalam, Manipuri, Marathi, Nepali, Odia, Sanskrit, Santali, Sindhi, Tamil, Telugu, and Urdu. Thanks to its better prompt tokenizer, it can easily be extended to other languages. This tokenizer has a larger vocabulary and handles byte…
VieNeu-TTS v3 Turbo is the next generation of Vietnamese TTS — 48 kHz high-fidelity speech, 23 built-in preset voices across three regions (North / Central / South), instant voice cloning, real-time streaming with an OpenAI-compatible API (16 concurrent streams on one RTX 3060), inline emotion cues, and seamless bilingual (En–Vi) code-switching. The reference implementation is the vieneu Python SDK (v3.7.1). Its minimal install is torch-free: on CPU everything runs on ONNX Runtime (PyTorch is never imported), and on a CUDA machine it auto-switches to the PyTorch engine with automatic batching and a continuous-batching stream scheduler — same API, no code change. The VieNeu-TTS v3 Turbo…