Text-to-speech, voice cloning, voice design, and vocoding with ONNX Runtime.
Export and run TTS models using text and optional prompt audio. The repository contains export targets across the model families below, each with Export, Inference, and Optimize scripts.
The shared model table lists public ONNX audio rates. Exporters handle native-rate conversion, while mode-dependent inputs are configured in the inference script.
The shared performance table reports RTF, calculated as synthesis time divided by audio duration. Lower is better. Unless noted, tests use a six-second reference and generate about 15 words. BigVGAN uses a (1, 100, 512) mel input.
IN_SAMPLE_RATE, OUT_SAMPLE_RATE, IN_AUDIO_DTYPE, and OUT_AUDIO_DTYPE in each exporter; KaniTTS and BigVGAN expose output settings only.F16, F32, and INT16; floating point uses [-1, 1], while INT16 uses PCM amplitude.torch.nn.functional.interpolate.Metadata.onnx stores the fixed package contract; runtime controls remain in the inference script. Inference validates metadata, graph layout, and referenced files.基于 ONNX Runtime 的语音合成、声音克隆、声音设计与神经声码器。
输入文本及可选提示音频即可导出并运行 TTS 模型。本项目包含下表所列模型家族的导出目标,每个目标均提供 Export、Inference 与 Optimize 脚本。
共用模型表列出公开 ONNX 音频采样率。导出脚本负责原生采样率转换,模式相关输入则在推理脚本中配置。
共用性能表中的 RTF 为合成耗时除以音频时长,数值越低越好。除特别注明外,测试使用六秒参考音频并生成约 15 个单词。BigVGAN 使用 (1, 100, 512) mel 输入。
IN_SAMPLE_RATE、OUT_SAMPLE_RATE、IN_AUDIO_DTYPE 与 OUT_AUDIO_DTYPE 设置公开音频输入输出;KaniTTS 与 BigVGAN 仅提供输出设置。F16、F32 与 INT16;浮点使用 [-1, 1],INT16 使用 PCM 幅值。torch.nn.functional.interpolate 在 ONNX 图内完成。Metadata.onnx 保存固定模型包约定,运行时配置保留在推理脚本中。推理时会校验元数据、图布局与引用文件。| Capability 能力 | Model 模型 | Audio I/O 音频输入输出 | Prompt Audio 提示音频 | Code 代码 | Source 来源 |
|---|---|---|---|---|---|
| Neural vocoder 神经声码器 | BigVGAN V2 | Mel → 24 kHz | No 否 | BigVGAN | GitHub |
| Voice cloning · voice design · voice direction 声音克隆 · 声音设计 · 语音指导 | Breeze TTS 2 | 24 → 24 kHz | Mode-dependent 取决于模式 | Breeze_TTS/v2 | GitHub |
| Voice cloning 声音克隆 | F5-TTS | 24 → 24 kHz | Required 必需 | F5_TTS | GitHub |
| Clone · voice design · speech editing 克隆 · 声音设计 · 语音编辑 | FireRedTTS3 | 24 → 24 kHz | Mode-dependent 取决于模式 | FireRedTTS/v3 | GitHub |
| Voice cloning 声音克隆 | IndexTTS 1.5 | 24 → 24 kHz | Supported 支持 | Index_TTS/v1.5 | GitHub |
| Voice cloning · emotion control 声音克隆 · 情感控制 | IndexTTS 2 | 22.05 → 22.05 kHz | Supported 支持 | Index_TTS/v2 | GitHub |
| Voice cloning · emotion control 声音克隆 · 情感控制 | IndexTTS 2.5 | 22.05 → 22.05 kHz | Supported 支持 | Index_TTS/v2 | GitHub |
| Text-to-speech 语音合成 | Inflect | Text → 24 kHz | No 否 | Inflect | GitHub |
| Text-to-speech 语音合成 | KaniTTS | Text → 22.05 kHz | No 否 | Kani_TTS | GitHub |
| Voice cloning · continuation 声音克隆 · 续写 | MOSS-TTS Nano | 48 → 48 kHz | Mode-dependent 取决于模式 | MOSS_TTS | GitHub |
| Clone · custom voice · voice design 克隆 · 定制音色 · 声音设计 | Qwen3-TTS | 24 → 24 kHz | Mode-dependent 取决于模式 | Qwen_TTS | GitHub |
| English voice cloning 英文声音克隆 | Raon-OpenTTS-0.3B | 16 → 16 kHz | Required 必需 | Raon_OpenTTS | GitHub |
| English voice cloning 英文声音克隆 | Raon-OpenTTS-1B | 16 → 16 kHz | Required 必需 | Raon_OpenTTS | GitHub |
| Voice cloning 声音克隆 | VoxCPM 1.5 | 44.1 → 44.1 kHz | Supported 支持 | VoxCPM/v1.5 | ModelScope |
| Clone · continuation · voice design 克隆 · 续写 · 声音设计 | VoxCPM 2 | 16 → 48 kHz | Mode-dependent 取决于模式 | VoxCPM/v2 | ModelScope |
| Voice cloning · dialogue 声音克隆 · 对话 | ZipVoice | 24 → 24 kHz | Required 必需 | ZipVoice | GitHub |
| OS 系统 | Device 设备 | Backend 后端 | Model 模型 | Precision 精度 | Time (s) 耗时(秒) | RTF |
|---|---|---|---|---|---|---|
| Ubuntu 24.04 | Intel Core i7-1165G7 | CPU | F5-TTS · NFE=32 | f32 | 180 | 60 |
| Ubuntu 24.04 | NVIDIA GeForce MX150 | GPU | F5-TTS · NFE=32 | f32 | 62 | 21 |
| Ubuntu 24.04 | Intel Core i7-1165G7 | CPU | IndexTTS | f32 | 18 | 6 |
| Ubuntu 24.04 | NVIDIA GeForce MX150 | GPU | BigVGAN V2 24khz_100band_256x | f16 | 4.6 | 1.53 |
| Ubuntu 24.04 | Intel Core i7-1165G7 | CPU | KaniTTS | q8f32 | 8.4 | 1.4 |
| Ubuntu 24.04 | Intel Core i7-1165G7 | CPU | KaniTTS | q4f32 | 5.2 | 0.87 |
| Ubuntu 24.04 | Intel Core i3-12300 | CPU | VoxCPM 1.5 | q8f32 | 9 | 1.5 |
| Ubuntu 24.04 | NVIDIA GeForce RTX 5060 Ti | GPU | VoxCPM 1.5 | f16 | 1.03 | 0.17 |
| Ubuntu 24.04 | Intel Core i3-12300 | CPU | Qwen3-TTS-0.6B-Base | q8f32 | 19 | 3.1 |
| Ubuntu 24.04 | Intel Core i3-12300 | CPU | VoxCPM 2 | q8f32 | 23 | 3.8 |
| Ubuntu 24.04 | NVIDIA GeForce RTX 5060 Ti | GPU | VoxCPM 2 | f16 | 2.05 | 0.34 |
| Ubuntu 24.04 | Intel Core i7-1165G7 | CPU | FireRedTTS3 | q8f32 | 31.1 | 5.2 |
353 commits
Python
100.0%
Text-to-speech, voice cloning, voice design, and vocoding with ONNX Runtime.
Export and run TTS models using text and optional prompt audio. The repository contains export targets across the model families below, each with Export, Inference, and Optimize scripts.
The shared model table lists public ONNX audio rates. Exporters handle native-rate conversion, while mode-dependent inputs are configured in the inference script.
The shared performance table reports RTF, calculated as synthesis time divided by audio duration. Lower is better. Unless noted, tests use a six-second reference and generate about 15 words. BigVGAN uses a (1, 100, 512) mel input.
IN_SAMPLE_RATE, OUT_SAMPLE_RATE, IN_AUDIO_DTYPE, and OUT_AUDIO_DTYPE in each exporter; KaniTTS and BigVGAN expose output settings only.F16, F32, and INT16; floating point uses [-1, 1], while INT16 uses PCM amplitude.torch.nn.functional.interpolate.Metadata.onnx stores the fixed package contract; runtime controls remain in the inference script. Inference validates metadata, graph layout, and referenced files.基于 ONNX Runtime 的语音合成、声音克隆、声音设计与神经声码器。
输入文本及可选提示音频即可导出并运行 TTS 模型。本项目包含下表所列模型家族的导出目标,每个目标均提供 Export、Inference 与 Optimize 脚本。
共用模型表列出公开 ONNX 音频采样率。导出脚本负责原生采样率转换,模式相关输入则在推理脚本中配置。
共用性能表中的 RTF 为合成耗时除以音频时长,数值越低越好。除特别注明外,测试使用六秒参考音频并生成约 15 个单词。BigVGAN 使用 (1, 100, 512) mel 输入。
IN_SAMPLE_RATE、OUT_SAMPLE_RATE、IN_AUDIO_DTYPE 与 OUT_AUDIO_DTYPE 设置公开音频输入输出;KaniTTS 与 BigVGAN 仅提供输出设置。F16、F32 与 INT16;浮点使用 [-1, 1],INT16 使用 PCM 幅值。torch.nn.functional.interpolate 在 ONNX 图内完成。Metadata.onnx 保存固定模型包约定,运行时配置保留在推理脚本中。推理时会校验元数据、图布局与引用文件。| Capability 能力 | Model 模型 | Audio I/O 音频输入输出 | Prompt Audio 提示音频 | Code 代码 | Source 来源 |
|---|---|---|---|---|---|
| Neural vocoder 神经声码器 | BigVGAN V2 | Mel → 24 kHz | No 否 | BigVGAN | GitHub |
| Voice cloning · voice design · voice direction 声音克隆 · 声音设计 · 语音指导 | Breeze TTS 2 | 24 → 24 kHz | Mode-dependent 取决于模式 | Breeze_TTS/v2 | GitHub |
| Voice cloning 声音克隆 | F5-TTS | 24 → 24 kHz | Required 必需 | F5_TTS | GitHub |
| Clone · voice design · speech editing 克隆 · 声音设计 · 语音编辑 | FireRedTTS3 | 24 → 24 kHz | Mode-dependent 取决于模式 | FireRedTTS/v3 | GitHub |
| Voice cloning 声音克隆 | IndexTTS 1.5 | 24 → 24 kHz | Supported 支持 | Index_TTS/v1.5 | GitHub |
| Voice cloning · emotion control 声音克隆 · 情感控制 | IndexTTS 2 | 22.05 → 22.05 kHz | Supported 支持 | Index_TTS/v2 | GitHub |
| Voice cloning · emotion control 声音克隆 · 情感控制 | IndexTTS 2.5 | 22.05 → 22.05 kHz | Supported 支持 | Index_TTS/v2 | GitHub |
| Text-to-speech 语音合成 | Inflect | Text → 24 kHz | No 否 | Inflect | GitHub |
| Text-to-speech 语音合成 | KaniTTS | Text → 22.05 kHz | No 否 | Kani_TTS | GitHub |
| Voice cloning · continuation 声音克隆 · 续写 | MOSS-TTS Nano | 48 → 48 kHz | Mode-dependent 取决于模式 | MOSS_TTS | GitHub |
| Clone · custom voice · voice design 克隆 · 定制音色 · 声音设计 | Qwen3-TTS | 24 → 24 kHz | Mode-dependent 取决于模式 | Qwen_TTS | GitHub |
| English voice cloning 英文声音克隆 | Raon-OpenTTS-0.3B | 16 → 16 kHz | Required 必需 | Raon_OpenTTS | GitHub |
| English voice cloning 英文声音克隆 | Raon-OpenTTS-1B | 16 → 16 kHz | Required 必需 | Raon_OpenTTS | GitHub |
| Voice cloning 声音克隆 | VoxCPM 1.5 | 44.1 → 44.1 kHz | Supported 支持 | VoxCPM/v1.5 | ModelScope |
| Clone · continuation · voice design 克隆 · 续写 · 声音设计 | VoxCPM 2 | 16 → 48 kHz | Mode-dependent 取决于模式 | VoxCPM/v2 | ModelScope |
| Voice cloning · dialogue 声音克隆 · 对话 | ZipVoice | 24 → 24 kHz | Required 必需 | ZipVoice | GitHub |
| OS 系统 | Device 设备 | Backend 后端 | Model 模型 | Precision 精度 | Time (s) 耗时(秒) | RTF |
|---|---|---|---|---|---|---|
| Ubuntu 24.04 | Intel Core i7-1165G7 | CPU | F5-TTS · NFE=32 | f32 | 180 | 60 |
| Ubuntu 24.04 | NVIDIA GeForce MX150 | GPU | F5-TTS · NFE=32 | f32 | 62 | 21 |
| Ubuntu 24.04 | Intel Core i7-1165G7 | CPU | IndexTTS | f32 | 18 | 6 |
| Ubuntu 24.04 | NVIDIA GeForce MX150 | GPU | BigVGAN V2 24khz_100band_256x | f16 | 4.6 | 1.53 |
| Ubuntu 24.04 | Intel Core i7-1165G7 | CPU | KaniTTS | q8f32 | 8.4 | 1.4 |
| Ubuntu 24.04 | Intel Core i7-1165G7 | CPU | KaniTTS | q4f32 | 5.2 | 0.87 |
| Ubuntu 24.04 | Intel Core i3-12300 | CPU | VoxCPM 1.5 | q8f32 | 9 | 1.5 |
| Ubuntu 24.04 | NVIDIA GeForce RTX 5060 Ti | GPU | VoxCPM 1.5 | f16 | 1.03 | 0.17 |
| Ubuntu 24.04 | Intel Core i3-12300 | CPU | Qwen3-TTS-0.6B-Base | q8f32 | 19 | 3.1 |
| Ubuntu 24.04 | Intel Core i3-12300 | CPU | VoxCPM 2 | q8f32 | 23 | 3.8 |
| Ubuntu 24.04 | NVIDIA GeForce RTX 5060 Ti | GPU | VoxCPM 2 | f16 | 2.05 | 0.34 |
| Ubuntu 24.04 | Intel Core i7-1165G7 | CPU | FireRedTTS3 | q8f32 | 31.1 | 5.2 |
353 commits
Python
100.0%