---
title: Irodori-TTS-500M-v3-fp16
canonical_url: "https://www.modelscope.cn/models/mlx-community/Irodori-TTS-500M-v3-fp16"
md_url: "https://www.modelscope.cn/models/mlx-community/Irodori-TTS-500M-v3-fp16.md"
repository: mlx-community/Irodori-TTS-500M-v3-fp16
last_updated: 2026-05-18
pipeline_tag: text-to-speech
tasks:
  - text-to-speech
model_type:
  - irodori_tts
base_model:
  - Aratako/Irodori-TTS-500M-v3
base_model_relation: finetune
parameters: 619.4M
tensor_type:
  - F32
  - F16
library_name:
  - mlx
  - safetensors
  - pytorch
frameworks:
  - pytorch
downloads: 10
stars: 0
tags:
  - mlx
  - japanese
---

# Irodori-TTS-500M-v3-fp16

> Irodori-TTS-500M-v3-fp16 - mlx-community 在 ModelScope 开源的模型。mlx-community/Irodori-TTS-500M-v3-fp16

mlx-community/Irodori-TTS-500M-v3-fp16 是 ModelScope 魔搭社区上的 619.4M 参数text-to-speech模型，基于 Aratako/Irodori-TTS-500M-v3 构建。

- **Repository**: mlx-community/Irodori-TTS-500M-v3-fp16
- **Tasks**: text-to-speech
- **Parameters**: 619.4M
- **Base model**: Aratako/Irodori-TTS-500M-v3
- **Tags**: mlx, japanese
- **Downloads**: 10
- **Stars**: 0
- **Last updated**: 2026-05-18

Source: https://www.modelscope.cn/models/mlx-community/Irodori-TTS-500M-v3-fp16

---

# mlx-community/Irodori-TTS-500M-v3-fp16

Japanese TTS model converted to MLX fp16 format from [Aratako/Irodori-TTS-500M-v3](https://huggingface.co/Aratako/Irodori-TTS-500M-v3) using [mlx-audio v0.4.3](https://github.com/Blaizzy/mlx-audio).

## Features (v3)

- **Duration predictor**: automatically estimates output length — no need to specify 
- **Sway Sampling**: non-linear timestep schedule for better quality with fewer steps
- **Voice cloning**: optional  for speaker style transfer
- **Backward compatible**: v2 checkpoints continue to work

## Available variants

| Model | Size | Memory |
|-------|------|--------|
| [mlx-community/Irodori-TTS-500M-v3-fp16](https://huggingface.co/mlx-community/Irodori-TTS-500M-v3-fp16) (this) | 977 MB | ~4.3 GB |
| [mlx-community/Irodori-TTS-500M-v3-8bit](https://huggingface.co/mlx-community/Irodori-TTS-500M-v3-8bit) | 663 MB | ~4.0 GB |

## Use with mlx-audio



### Command line



### Python