Model reference · open weights
IndexTTS-2.5-Comfy is an open-weight audio or speech model from t8star. AxForge deploys and operates it for you on dedicated EU-owned hardware — with the licence handled where one is required.
Available as managed deployment — configured and operated for you on dedicated EU hardware, quoted per deployment.
What it is
| Released by | t8star |
|---|---|
| Type | Audio & music |
| Task | Text→speech |
| Runs with | indextts |
| Released | 2026-08-27 |
| Popularity | 533 downloads / month |
| Licence | Commercial licence needed |
About
这是供以下两个项目共用的 IndexTTS 2.5 完整推理模型包:
桌面版和 ComfyUI 使用的是同一套模型文件,无需分别下载。桌面启动器可以直接选择 ComfyUI 的
models/TTS/IndexTTS-2.5 目录,共享一份约 7.72 GiB 的模型。
本仓库解决官方 IndexTeam/IndexTTS-2.5 模型仓库未包含 bpe.model,以及首次运行还会从多个
仓库下载 Wav2Vec2-BERT、CAMPPlus 和 BigVGAN 的问题。这里把实际推理需要的 26 个文件集中在一个
目录中,方便自动下载、断点续传、离线运行和 SHA-256 校验。
本仓库不是 IndexTTS 官方发布渠道。所有权重保持源文件内容不变,仅重新整理目录。 原始权利人不对此第三方整理仓库及其衍生集成提供认可、担保或保证,并对其不承担责任。
仓库根目录的 model-bundle.json 是机器可读的完整文件清单,model-bundle.sig 是对应的
Ed25519 签名。T8star-Aix 桌面启动器会先验证签名,再信任清单中的仓库版本、文件路径、大小和
SHA-256;只下载缺失或损坏的文件,网络中断后可以继续下载。
模型更新由 bundleVersion 判断,不使用仓库 HEAD 判断。因此仅修改 README 不会误报模型更新。
程序更新仍通过 GitHub Release 独立发布,模型不会被打进程序增量包。
节点代码和模型是两个不同目录:
ComfyUI/
├─ custom_nodes/
│ └─ comfyui-indextts25-T8/ ← GitHub 节点代码
└─ models/
└─ TTS/
└─ IndexTTS-2.5/ ← 本 HF 仓库的模型文件
├─ config.yaml
├─ bpe.model
├─ gpt.pth
├─ codec.pth
├─ s2mel.pth
├─ feat1.pt
├─ feat2.pt
├─ wav2vec2bert_stats.pt
├─ multilingual_zh_ja_yue_char_del.tiktoken
├─ qwen0.6bemo4-merge/
└─ hf_cache/
├─ w2v-bert-2.0/
├─ campplus_cn_common.bin
└─ bigvgan/
Windows 完整路径示例:
D:\ComfyUI\models\TTS\IndexTTS-2.5\config.yaml
D:\ComfyUI\models\TTS\IndexTTS-2.5\bpe.model
D:\ComfyUI\models\TTS\IndexTTS-2.5\hf_cache\bigvgan\bigvgan_generator.pt
请确认 config.yaml 和 bpe.model 是 IndexTTS-2.5 的直接子文件。不要多套一层:
ComfyUI/models/TTS/IndexTTS-2.5/IndexTTS-2.5-Comfy/config.yaml ← 错误
桌面启动器支持两种方式:
IndexTTS-2.5;ComfyUI/models/TTS/IndexTTS-2.5。模型、保存音色、生成音频和用户设置都独立于桌面程序更新,更新应用不会删除这些内容。
pip install -U huggingface_hub
hf download t8star/IndexTTS-2.5-Comfy --local-dir "ComfyUI/models/TTS/IndexTTS-2.5"
也可以在仓库页面选择 Files and versions → Download。解压后确保 config.yaml、bpe.model、
gpt.pth 和 hf_cache 位于同一个 IndexTTS-2.5 目录。
bpe.modelhf_cache/w2v-bert-2.0:Transformers 权重、配置和预处理配置hf_cache/campplus_cn_common.bin:音色特征模型hf_cache/bigvgan:22.05 kHz BigVGAN 配置和生成器权重没有收录训练优化器、判别器、重复的 Fairseq conformer_shaw.pt 或当前 IndexTTS 2.5 推理代码
不会读取的 MaskGCT 独立权重。
| 内容 | 来源 | 固定版本 |
|---|---|---|
| IndexTTS 2.5 主模型 | IndexTeam/IndexTTS-2.5 | c39ce5ba981572cb187443877ff559dfb246ce63 |
bpe.model | IndexTeam/IndexTTS-2 | 740dcaff396282ffb241903d150ac011cd4b1ede |
| Wav2Vec2-BERT 2.0 | facebook/w2v-bert-2.0 | da985ba0987f70aaeb84a80f2851cfac8c697a7b |
| CAMPPlus | funasr/campplus | e4b6ede7ce16997aff4ae69fbca1f0175e2afede |
| BigVGAN | nvidia/bigvgan_v2_22khz_80band_256x | 633ff708ed5b74903e86ff1298cf4a98e921c513 |
详细归属和许可证见 THIRD_PARTY_NOTICES.md。下载或使用即表示你已阅读并
接受本仓库 LICENSE 中的 bilibili Model Use License Agreement 及各第三方模型许可证。
This is the shared, inference-only IndexTTS 2.5 model bundle for the T8star-Aix Desktop package and
ComfyUI node. Both applications use the same byte-identical files; the Desktop launcher can directly
select ComfyUI/models/TTS/IndexTTS-2.5, so a second copy is unnecessary.
The repository combines the pinned official IndexTTS 2.5 files, the required bpe.model, and the exact
Wav2Vec2-BERT, CAMPPlus, and BigVGAN runtime files. model-bundle.json and model-bundle.sig provide a
signed, machine-readable version and per-file SHA-256 list for resumable automatic download and repair.
This is an unofficial packaging mirror and is not endorsed, warranted, or guaranteed by the original
right-holders. Review LICENSE and THIRD_PARTY_NOTICES.md before downloading or using the files.
From the published model card. Full card on the HuggingFace links in the sidebar.
How it works
Using it via the API
Once AxForge deploys indextts-2-5-comfy for you, it answers on the OpenAI-compatible API — the same base URL and keys as every other model. (indextts-2-5-comfy below is illustrative; you get the exact model name on deployment.)
$ curl -sS https://api.axforge.ai/v1/audio/transcriptions \ -H "Authorization: Bearer $AXFORGE_API_KEY" \ -F model="indextts-2-5-comfy" -F file=@audio.mp3
Create an account — your API key is available in the console. 3M free tokens every 30 days with every new account.