Overview
- Full name
docker.io/jonathanwx/cosyvoice2-vllm- Registry
- docker.io
- Namespace
- jonathanwx
- Repository
- cosyvoice2-vllm
- Cached tags
- 4
- Last synced
- 07/06/2026, 08:42 AM
- Topics
Introduction
CosyVoice2 vLLM TTS 推理镜像，基于 vLLM 加速多语言语音合成 API，适合低延迟 TTS 服务 GPU 部署。
Details
jonathanwx/cosyvoice2-vllm 集成 CosyVoice2 与 vLLM serving，HTTP API 文本转语音。定位为 GPU TTS 推理服务；K8s Deployment 需 GPU 节点。
快速启动
docker run -d --name cosyvoice2 \
-p 8000:8000 \
--gpus all \
docker.io/jonathanwx/cosyvoice2-vllm:latest
推荐实践
docker run -d --name cosyvoice2 \
--restart unless-stopped \
-p 8000:8000 \
--gpus all \
-e MODEL=cosyvoice2 \
docker.io/jonathanwx/cosyvoice2-vllm:1.0.0
核心参数说明
-p 8000:8000— TTS API 端口--gpus all— vLLM GPU 推理-e MODEL— 模型名称- 固定 tag — CosyVoice2/vLLM pin
Kubernetes
apiVersion: apps/v1
kind: Deployment
metadata:
name: cosyvoice2
spec:
replicas: 2
selector:
matchLabels:
app: cosyvoice2
template:
metadata:
labels:
app: cosyvoice2
spec:
containers:
- name: tts
image: docker.io/jonathanwx/cosyvoice2-vllm:1.0.0
ports:
- containerPort: 8000
resources:
limits:
nvidia.com/gpu: "1"
---
apiVersion: v1
kind: Service
metadata:
name: cosyvoice2
spec:
selector:
app: cosyvoice2
ports:
- port: 8000
targetPort: 8000
GPU 节点池；按 RTF 调副本与显存。
Latest tags (20)
View all 4 tags →| Tag | Arch | Pushed At | Size | Digest | Sync | Wish | Action |
|---|---|---|---|---|---|---|---|
docker.io/jonathanwx/cosyvoice2-vllm :0.4-cuda128 | linux/amd64 | 2026/01/29 19:40 | 14.7 GB | sha256:0ce79e311fc9… | |||
docker.io/jonathanwx/cosyvoice2-vllm :0.3 | linux/amd64 | 2025/09/27 19:12 | 14.2 GB | sha256:74bc73b7406b… | |||
docker.io/jonathanwx/cosyvoice2-vllm :0.2 | linux/amd64 | 2025/09/09 00:51 | 10.6 GB | sha256:9221d87bb643… | |||
docker.io/jonathanwx/cosyvoice2-vllm :0.1 | linux/amd64 | 2025/09/07 20:14 | 11.6 GB | sha256:75df99dc9903… |