SemBridge bridges discrete semantic supervision and continuous acoustic generation. A shared semantic tokenizer anchors both acoustic latents and autoregressive language-model states, improving linguistic fidelity while retaining continuous-latent generation. The unified model supports text-to-speech, note-conditioned singing, and mixed speech and singing. - [2026-10]: Released the inference code and model checkpoints. - [2026-08]: Released the arXiv paper. Use Python 3.10. The repository includes 10 complete examples from our Demo page: 4 TTS, 4 SVS, and 2 Unified cases, with reference audio and prepared requests. SVS and Unified include aligned lyrics, pitches, durations, and…
Open-weight model · Text to speech
Irodori-TTS-v4.1-Small-fp32
by HIDING FOX hidingfox/Irodori-TTS-v4.1-Small-fp32
Irodori-TTS-v4.1-Small-fp32 is an open-weight model for text to speech from HIDING FOX. It has 766M parameters. At 16-bit it needs about 1.8 GB of GPU memory, which fits on 1x MI300X from $1.85 an hour, at the lowest prices in the SAVRN Index.
This model was converted to MLX format from Aratako/Irodori-TTS-v4.1-Small using mlx-audio version 0.4.7. prediction in one checkpoint.
Parameters766M
Context—
Weights3.5 GB
License—
AccessOpen weights
Monthly Downloads—
Runs On
What it takes to serve Irodori-TTS-v4.1-Small-fp32 (766M 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 | 1.5 GB | 1.8 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
| 8-bit | 0.8 GB | 0.9 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
| 4-bit | 0.4 GB | 0.5 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 Oct 18, 2026.
Irodori-TTS-v4.1-Small-fp32 on every accelerator the SAVRN Index prices, at every precision
Model Card
This model was converted to MLX format from Aratako/Irodori-TTS-v4.1-Small using mlx-audio version 0.4.7. prediction in one checkpoint. - v4.1 differs from v4 only in the duration predictor, which upstream retrained separately with every other parameter frozen. The remaining 683 of 714 tensors are bit-identical to v4. - The ModernBERT-ja-310m text encoder weights and its tokenizer are bundled, and the Semantic-DACVAE-Japanese-32dim codec ships in dacvae/, so inference needs no additional downloads. and concatenates them, which matches training better than one long recording. - Short caption-only prompts are shorter than in v4, but still over-predicted. Measured with the same reference clip…
Excerpt from the card by HIDING FOX.
Configuration
- Model type
- irodori_tts
Identity and Version
- Repository
- hidingfox/Irodori-TTS-v4.1-Small-fp32
- Publisher
- HIDING FOX
- Task
- Text to speech
- Modality
- Audio
- Library
- mlx-audio
- Parameters
- 766M parameters
- Languages
- mlx-audio, mlx
- Revision
- b7d1b8402f945e007549be71b201a186d5b24b09
- First published
- 2026-09-21
- Last updated
- 2026-09-21
Files and Weights
8 files, 3.5 GB in total. The weights are 2 files totalling 3.5 GB in safetensors.
Weights2 files · 3.5 GB
Configuration2 files · 5.0 KB
Tokenizer2 files · 6.7 MB
Documentation1 file · 2.3 KB
Repository1 file · 1.5 KB
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| dacvae/model.safetensors | Weights | 429.5 MB | 0bf3ff42e810 |
| model.safetensors | Weights | 3.1 GB | 40f6e8317bab |
| config.json | Configuration | 4.6 KB | — |
| dacvae/config.json | Configuration | 329 B | — |
| README.md | Documentation | 2.3 KB | — |
| .gitattributes | Repository | 1.5 KB | — |
| tokenizer/tokenizer.json | Tokenizer | 6.7 MB | — |
| tokenizer/tokenizer_config.json | Tokenizer | 668 B | — |
License and Download
- License
- Not stated by the source
- Access
- Open weights, no gate
- Download size
- 3.5 GB
Download from HIDING FOX
Released by HIDING FOX through its official repository on Hugging Face.
Built From
- Derived from Aratako/Irodori-TTS-v4.1-Small
Memory Requirements
| Precision | Weights in memory |
|---|---|
| As published | 3.5 GB |
| 16-bit | 1.5 GB |
| 8-bit | 0.8 GB |
| 4-bit | 0.4 GB |
Weights only, from the published parameter count; the key-value cache and runtime add to this.
Questions About Irodori-TTS-v4.1-Small-fp32
How much GPU memory does Irodori-TTS-v4.1-Small-fp32 need?
About 1.8 GB at 16-bit and 0.5 GB at 4-bit: the weights (766M parameters) plus a working margin. A long context needs more.
What is the cheapest GPU to run Irodori-TTS-v4.1-Small-fp32 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.
Similar Models
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
This Vietnamese text-to-speech model is fine-tuned from g-group-ai-lab/gwen-tts-0.6B, based on the Qwen3-TTS architecture. The exported package includes merged weights and the ngochuyen speaker in CustomVoice mode. No reference audio is required at inference time. The model was fine-tuned on approximately 90 hours of Ngọc Huyền's voice, collected by crawling YouTube and TikTok. Training used LoRA adaptation, adapter merging, and selective refinement of the talker and output head, followed by joint refinement with the code predictor. The final model was selected by the training pipeline and packaged with a speaker embedding derived from reference audio. If the repository is private…
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…
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…
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…