SAVRN Model Hub · Models by Task
Text to Speech Models
148 open-weight text to speech models in the SAVRN Model Hub, with X Girls, Qwen and OpenMOSS publishing the most.
148Models
106Publishers
10M to 128.2BParameter range
8Licenses
SAVRN's Take
Nobody buys a rack for text to speech. The 31 models on this page with a parameter count run from a 10 million parameter voice tuner for Kokoro to OpenMOSS's 8.5 billion parameter MOSS-TTS-v1.5, which wants 20.4 GB at 16-bit and 5.1 GB at 4-bit. Qwen's 1.9 billion parameter Qwen3-TTS-12Hz-1.7B-CustomVoice fits in 4.6 GB. Every indexed entry prices out at one MI300X at $1.85 an hour, so the GPU is not where the cost lives. What differs is the job: XTTS-v2 clones a voice from a 6-second clip across 17 languages, Qwen3-TTS takes tone and speaking rate instructions across 10, OmniVoice lists more than 600, and MOSS-TTS-v1.5 pairs zero-shot cloning with long-form generation on a 40,960 token context.
Downloads favor the small end. Hexgrad's Kokoro-82M, 82 million parameters under Apache 2.0, pulled 11,687,399 in a month, ahead of Coqui's XTTS-v2 at 7,324,519 and audio.cpp-gguf at 4,129,715, GGUF conversions for that runtime rather than a model. Qwen, OpenMOSS and AI at Meta hold four entries each.
Licensing is where we would spend the diligence. Apache 2.0 covers 32 of the 64 and MIT another 10. Nine carry CC BY-NC 4.0, F5-TTS and its 984,345 downloads among them, and the NC means noncommercial. Seven are filed as other, XTTS-v2 and audio.cpp-gguf included, two state no license, and OmniVoice's field is blank. Read the license text on the publisher's repository for anything outside Apache and MIT, and pick the runtime before the model: coqui, chatterbox, f5-tts, omnivoice and audio.cpp are each their own library, and that choice outlasts any checkpoint.
Most Downloaded
| Model | Publisher | Parameters | License | Monthly downloads | Cheapest GPUs at 16-bit |
|---|---|---|---|---|---|
| Kokoro-82M | Hexgrad | — | apache-2.0 | 10.1M | — |
| audio.cpp-gguf | Audio.cpp | — | other | 6.6M | — |
| XTTS-v2 | Coqui.ai | — | other | 6.1M | — |
| OmniVoice | K2 FSA | 613M | Not stated | 2M | 1x MI300X, $1.85/hr |
| Qwen3-TTS-12Hz-1.7B-CustomVoice | Qwen | 1.9B | apache-2.0 | 1.9M | 1x MI300X, $1.85/hr |
| chatterbox | Resemble AI | — | mit | 1.8M | — |
| VieNeu-TTS-v3-Turbo | Pham Nguyen Ngoc Bao | 131M | apache-2.0 | 1.4M | 1x MI300X, $1.85/hr |
| kokoro-inno-clone-tuner | Jeremy Braun | 10M | apache-2.0 | 1.2M | 1x MI300X, $1.85/hr |
| Qwen3-TTS-12Hz-0.6B-CustomVoice | Qwen | 906M | apache-2.0 | 1M | 1x MI300X, $1.85/hr |
| F5-TTS | Yushen CHEN | — | cc-by-nc-4.0 | 984.3k | — |
Licenses
| License | Models | Commercial use |
|---|---|---|
| apache-2.0 | 72 | Yes |
| mit | 22 | Yes |
| other | 21 | Read the license |
| cc-by-nc-4.0 | 13 | Not without separate permission |
| not stated | 9 | Not stated |
| cc-by-4.0 | 9 | Yes |
| openrail++ | 1 | Yes, with conditions |
| gpl-3.0 | 1 | Read the license |
Who Publishes Them
| Publisher | Models |
|---|---|
| X Girls | 5 |
| Qwen | 4 |
| OpenMOSS | 4 |
| AI at Meta | 4 |
| Josh Leverette | 4 |
| Richard R. | 3 |
All 148 Models, Page 3 of 3
RKNN model artifacts and a checksum-pinned weights index for nanobot-channel-voice. The repository includes four official openWakeWord phrase heads, each available as an original-ONNX package for CPU/ONNX-capable hosts and as a validated RV1126B hybrid package: Choose one variant for the target host. Every fetch is a complete package: ONNX entries install the three original upstream graphs plus a phrase sidecar; RV1126B entries install the phrase head, exact Python-mel filterbank, and static RKNN embedding. The runtime fetches a complete model entry through nanobot-voice: weights-index.json is the authoritative manifest of available keys, source provenance, checksums, and acceptance…
Distribución de Neo TTS Studio 1.2 para Linux y Windows, en CPU y CUDA. Los cuatro paquetes son archivos individuales completos, sin partes. Incluyen motor de entrenamiento e inferencia C++17, exportación y panel Axolotl. No contienen modelos de voz privados, credenciales ni datasets. Comprueba las sumas de SHA256SUMS.txt después de descargar y antes de ejecutar. La edición CUDA necesita GPU NVIDIA y controladores compatibles; CUDA Windows está compilada con LibTorch CUDA 12.8, pero aún no se ha probado en un equipo Windows con NVIDIA. En Windows abre PowerShell en la carpeta del ZIP extraído:./neo-tts.exe --webui --port 8787. En Linux ejecuta./run.sh --webui --port 8787 --data…
Single-file checkpoints of k2-fsa/OmniVoice (Qwen3 LLM + HiggsAudioV2 codec), repacked for ComfyUI-OmniVoice-Native. Config, codec config, and BPE tokenizer are embedded as safetensors metadata — no sidecar files needed. - omnivoice.safetensors (3.3GB, full quality) — most robust, recommended for tags + voice-design prompts. - omnivoice-int8.safetensors (1.1GB, INT8-ConvRot) — same quality on plain prompts, smaller download. ~86% top-1 agreement with full. 1. Install the pack: ComfyUI Manager → search OmniVoice, or git clone https://github.com/Aero-Ex/ComfyUI-OmniVoice into ComfyUI/customnodes/. No pip install needed. 2. Put a file from this repo into ComfyUI/models/omnivoice/. 3. Load…
Self-contained ONNX export of k2-fsa/OmniVoice for CPU inference, including voice cloning: a dynamic-INT8 Qwen3-0.6B diffusion backbone, the Higgs audio-v2 encoder and vocoder in FP32, and the tokenizer/config files. Nothing is downloaded from any other model repo. Backbone: inputids[B,8,S] int64, audiomask[B,S] bool, attentionmask[B,1,S,S] bool → logits[B,8,S,1025]. Encoder: wav[1,1,N] float32 (24 kHz, N a multiple of 960) → audiocodes[1,8,N/960] int64. Vocoder: audiocodes[B,8,T] int64 → audiovalues[B,1,960T] (24 kHz). B, S, N and T are dynamic. --threads should be the physical core count. The omnivoice package supplies only the diffusion loop and text front-end code; every weight comes…
Re-exported llmdecoder sub-model for OmniVoice Web, a fully in-browser (WebGPU) TypeScript runtime for k2-fsa/OmniVoice. The public onnx-community/OmniVoice-Onnx llmdecoder uses causal attention (GroupQueryAttention + 2-D causal mask). OmniVoice is a bidirectional masked-diffusion LM, so with causal attention the model ignores the text prompt and emits generic speech. This repo hosts the Qwen3-0.6B core re-exported with a 4-D all-zero (non-causal) attention mask, which restores exact text conditioning. All three reproduce the PyTorch reference with WER 0 (round-trip Whisper) on the evaluation sentence. The embeddings/heads/tokenizer/codec sub-models remain on tools/exportllm.py --dtype…
opedromartins/ASR-datasets-ptbr - freds0/TAGARELA basemodel: - canopylabs/orpheus-3b-0.1-ft pipelinetag: text-to-speech - text-to-speech - tts - orpheus - lora - unsloth - snac - portuguese - brazilian-portuguese - llama.cpp
Mixed-precision conversion of Paradee-8M v1.0, a single-voice English distillation of Kokoro. It accepts phonemes and generates mono 24 kHz Float PCM with its learned afheart voice. Glade is a native Swift library for speech inference on Apple silicon, using Core AI, Core ML, and model-specific host implementations. Its runtime source is currently private. This repository distributes the model assets and runtime metadata used by that library. Python is export/measurement tooling, not an application dependency. Mac qualification is included; iPhone and Watch qualification is not claimed. Download weights and configuration from one snapshot. INT8 text/prosody/acoustic weights use FP16…
[Español abajo / Spanish below] This model has been automatically converted from Piper Train checkpoints to ONNX format using Piper ONNX Converter. This model is optimized for use with Piper TTS. For Windows users, we recommend using ONNX TTS by HirCoir - a user-friendly GUI application: - Easy-to-use interface for text-to-speech conversion - Optimized for ONNX models like this one - esMX-capybara.onnx: Main ONNX model file - esMX-capybara.onnx.json: Model configuration file Este modelo ha sido convertido automáticamente de checkpoints de Piper Train a formato ONNX usando Piper ONNX Converter. Este modelo está optimizado para usar con Piper TTS. Para usuarios de Windows, recomendamos usar…
Specialized conversion of Kyutai Pocket TTS. This profile supports English text with the predefined alba voice and returns mono 24 kHz Float PCM, including incremental audio. Glade is a native Swift library for speech inference on Apple silicon, using Core AI, Core ML, and model-specific host implementations. Its runtime source is currently private. This repository distributes the model assets and runtime metadata used by that library. Python is export/measurement tooling, not an application dependency. Mac qualification is included; iPhone and Watch qualification is not claimed. Download weights and configuration from one snapshot. The approximately 117.2 MB bundle uses W8A16 language…
Optimized ONNX exports of Pocket TTS for high-speed, memory-efficient text-to-speech and zero-shot voice cloning on CPU, mobile, and edge runtimes. Mirrored and validated from KevinAHM/pocket-tts-onnx at commit 58a6d00cf13d239b6748cb0769f35c580a8f606c. This repository provides comprehensive support for both compact 6-layer models and high-capacity 24-layer models: - 6-Layer Compact Models: english2026-04, german, italian, portuguese, spanish - Ideal for lightweight CPU inference, mobile devices, and low-latency streaming. - ~302 MB FP32 / ~76 MB INT8 Flow LM. - 24-Layer Capacity Models: french24l, german24l, italian24l, portuguese24l, spanish24l - High-fidelity modeling for complex prosody…
中文语音合成(TTS)模型,支持语音克隆与数字/中英混读,CPU 优先设计,权重原生 int8 量化。 基于 Kyutai Pocket TTS 架构:6 层 FlowLM(dmodel 1024)+ Mimi 神经音频编解码器(24 kHz),端到端流式推理。本版本(R21)相对上一代的核心变化: - 原生 int8 交付:FlowLM 注意力与 FFN 权重已离线量化打包,无需运行时转换,体积 223 MB,CPU 速度提升约 1 倍 - 9174 token 词表:官方英文 BPE-4000 基底 + 追加单字汉字/全角标点,bytefallback 兜底生僻字符 - 中英混读:APP、VIP、5G、WiFi 等英文术语按中文习惯读法 - 模型为预量化 int8 权重(modelint8.pt),加载即用,无需 quantize=True 或有界面演示(需 pip install gradio): - 采样温度 temp=0.3(默认);文本越长对参考音频质量越敏感 文本准确性用 whisper-small 对合成音频转写后计算字错误率(CER),每个场景 10 次独立采样取均值,生产配置(单步采样)。数值越低越好。 测试环境:Intel i7-12700KF,生成 52 字客服长句(约 8.5 秒音频),每档 3 次取均值。 参考音频:客服目标音色(kefulong / digitsorder / codeswitch / shortunseen)与 AISHELL-3 说话人 A/B(speakerAlong /…
This is a Core ML build of Qwen/Qwen3-TTS-12Hz-0.6B-Base, the voice-cloning model. It clones a voice from a few seconds of reference audio plus its transcript, and runs it on the Neural Engine of an iPhone, iPad or Mac. It needs no GPU, MLX or ONNX Runtime. It's built for Gloam Voice Studio and its radio app by Tiny Trash Labs. The Swift runtime that loads it is the open-source QwenANE target in (see Sources/QwenANE/README.md). Requires iOS 18 / macOS 15 or later (stateful Core ML models). The set is about 1.9 GB. - Base, not CustomVoice. It clones any voice rather than offering preset speakers, and both encoders are included so a new voice can be prepared on the device. - One read is one…
First-party Nile Academy conversion of the official Qwen checkpoint at fd4b254389122332181a7c3db7f27e918eec64e3. This is a derivative ONNX artifact, not an ONNX release published by Qwen. It contains data only; Python is not required by BookAlive. The fp32 pack includes a speaker encoder, talker with an explicit KV cache, a complete code-predictor frame, speech decoder, embedding tables, tokenizer and publisher licence. BookAlive reuses its Qwen VoiceDesign generation implementation for the Base talker, predictor and decoder. The source implementation is pinned at 022e286b98fbec7e1e916cb940cdf532cd9f488e. The current adapter uses x-vector-only reference cloning. It supports ten languages…
本仓库提供由 Qwen/Qwen3-TTS-12Hz-0.6B-Base 转换得到的 RKNN TTS 模型。 - Model ID：RKNNAI/RK182X-TTS-Qwen3-TTS-12Hz-0.6B-Base - 模型显示名称：RK182X-TTS-Qwen3-TTS-12Hz-0.6B-Base - 源模型：Qwen/Qwen3-TTS-12Hz-0.6B-Base - 模型类型：TTS ModelScope 完整下载： Hugging Face 完整下载： ModelScope 指定配置下载： Hugging Face 指定配置下载： - 使用配套 RKNN Runtime 和驱动；运行前用 rknn-smi -v 检查设备端版本。 源模型采用 Apache License 2.0，正文见 LICENSE，归属和转换说明见 NOTICE。同时遵守 RKNN Toolkit 和 RKNN Runtime 相关许可。
本仓库提供由 Qwen/Qwen3-TTS-12Hz-1.7B-Base 转换得到的 RKNN TTS 模型。 - Model ID：RKNNAI/RK182X-TTS-Qwen3-TTS-12Hz-1.7B-Base - 模型显示名称：RK182X-TTS-Qwen3-TTS-12Hz-1.7B-Base - 源模型：Qwen/Qwen3-TTS-12Hz-1.7B-Base - 模型类型：TTS ModelScope 完整下载： Hugging Face 完整下载： ModelScope 指定配置下载： Hugging Face 指定配置下载： - 使用配套 RKNN Runtime 和驱动；运行前用 rknn-smi -v 检查设备端版本。 源模型许可证见根目录 LICENSE；同时遵守 RKNN Toolkit 和 RKNN Runtime 许可条款。
Native routed LoRA weights from omnicommon400/checkpoint-200. The tensor file is copied unchanged from the requested checkpoint. Common400 are in-sample and must not be presented as held-out performance. It is separate from the paper's cubic S→R main-result checkpoints; the paper's 6.85% CER and 4.51 MOS are not asserted for this artifact. Use Python 3.12 with a compatible PyTorch installation. Install the upstream OmniVoice 0.1.5 inference package and Hugging Face client: Replace the text and reference inputs with your own. The reference transcript must match the supplied recording. Use --device cpu if CUDA is unavailable. Use --base-dir /path/to/OmniVoice to reuse a local base-model…
A Santali speech-synthesis model for text written in the Ol Chiki script. The model was trained using the Piper VITS training pipeline and exported to ONNX for offline inference and deployment on supported platforms. The exported ONNX inference model. This is the primary deployment artifact. The corresponding Piper model configuration. Training configuration containing the model and text-symbol configuration. The custom text-symbol inventory used for Santali written in Ol Chiki. Training metadata for the speech samples used in this training run. Machine-readable metadata describing the model, training run, dataset, and validation results. The PyTorch Lightning checkpoint is retained…
A small English text-to-speech model that you steer with numbers. Each request can set any subset of 12 acoustic controls, for example speaking rate, loudness, median pitch, pitch range, final rise or fall, pauses, spectral tilt and breathiness. Each control is a z-score relative to the chosen voice's own neutral speech. You can also pick a style tag, give a free-text direction (which a small adapter turns into a request you can inspect), or use inline markup. The model can clone a voice from a short reference clip. The CLI looks for the model in the current directory. From anywhere else, pass it before the command (steertts --model WhissleAI/steer-tts say "Hi." -o hi.wav) or set…
Three single-speaker Vietnamese narrator voices in Piper/VITS format (22,050 Hz, ~63 MB each), ready for sherpa-onnx: the.onnx files carry sherpa-onnx metadata and each folder ships a tokens.txt. Phonemization uses eSpeak NG (vi); espeak-ng-data is not included. None of the voices belongs to a real person: each was designed from a text description with VoxCPM2 Voice Design. 1. Training data. For each voice, roughly 5,000 Vietnamese sentences were synthesized with VoxCPM2 (Apache-2.0). About 30% of the sentences contain English names and terms, so the voices learn eSpeak's English phonemes. Every clip was transcribed back with Whisper, and clips with a high character error rate, an…
The voice of an AI, written down and checkable. voaice is the voice stack of mindX: speech in, speech out, measured throughout. It gives a machine a voice, says the realm's names the way their owners say them, and proves which voice made a recording. This repository is voaice's voice library: 70 open-licensed Piper voices in 29 language families (4.62 GB), kept here so a mindX node holds only the voices it speaks with and fetches any other on demand. The voices in piper/ are here for anyone to use: download one, run it with Piper, build on it. Each voice keeps its own open licence, the licence of the dataset its speakers recorded, and the folder's official licences list every voice under…
VoxCPM2 is a tokenizer-free, diffusion autoregressive Text-to-Speech model — 2B parameters, 30 languages, 48kHz audio output, trained on over 2 million hours of multilingual speech data. - 30-Language Multilingual — No language tag needed; input text in any supported language directly - Voice Design — Generate a novel voice from a natural-language description alone (gender, age, tone, emotion, pace…); no reference audio required - Controllable Cloning — Clone any voice from a short clip, with optional style guidance to steer emotion, pace, and expression while preserving timbre - Ultimate Cloning — Provide reference audio + its transcript for audio-continuation cloning; every vocal nuance…
An INT8-quantized repack of openbmb/VoxCPM2, prepared so the model can be loaded and run entirely inside a web browser by shi3z/voxcpm2web (a WASM + WebGPU port of Nothing about the model architecture was changed. This repository exists to solve 1. wasm32 has a 4 GB address space. The upstream checkpoint is 4.27 GiB of weights plus a 359 MB audiovae.pth. 2. Quantizing during load costs VRAM you may not have. Doing it ahead of time means the browser never materializes the full-precision tensors at all. There is no separate audiovae file: the AudioVAE weights are already merged in, so a browser needs exactly one large download instead of two. Both rows produce applied=632 missing=0 unused=0…
Not a stock voice. A voice. A Chinese-exclusive TTS voice model by Zixi Li. VITS2 lineage (forked from MeloTTS), 51.43M parameters, Chinese-only, no BERT. - Small — 51M params is light enough for local inference, CPU included. No cloud, no API key, no GPU bill. - Chinese-exclusive — the whole stack serves one language: 77 symbols, 6 tones, zero language-routing overhead. - One real person's voice — a single real human's voice, so the output keeps one person's timbre, rhythm and breath. ▶ finalshowcase.wav — one take, default inference parameters (sdpratio=0.2, noisescale=0.6, noisescalew=0.8), text from Zixi Li's own fiction. Weights only — training data is not part of this release. The…
Questions
Which Text to speech models are most downloaded?
By monthly downloads reported by the Hugging Face Hub: .