Total runs:
935
Run Growth:
-1.1K
Growth Rate:
-118.07%
Sponsored by
AdsCreator.com
AXERA-TECH / AX-Meeting-Studio
Model's Last Updated: August 07 2026
audio-text-to-text
Introduction of AX-Meeting-Studio
Model Details of AX-Meeting-Studio
AX Meeting Studio
本仓库是面向 AX8850N 板端部署 的 AX Meeting Studio，用于将已经构建并验证的 wheel、外置模型、HTTPS 证书和启动脚本整体部署到设备上运行。
工程能力
AX Meeting Studio 面向离线和实时会议场景，提供语音活动检测、说话人聚类与归属、语音识别、 会议转写以及大模型总结工作台：
- **VAD + CAM++**：检测有效语音并进行说话人嵌入、聚类和标签归属。
-
双 ASR 后端
：默认的
sensevoice支持常规转写和词级时间戳；firered_punc使用 FireRed AED 加 CT-Transformer 标点，适合更高准确度的最终稿，但不提供词级时间戳。 -
多种使用方式
：提供 HTTPS Web UI、
ax-meeting-studio命令行和 Python API；Web UI 支持离线文件上传、浏览器麦克风实时转写和可选的大模型会议总结。 - 板端资源管理 ：VAD 与 CAM++ 常驻；两种 ASR 按需加载和切换，避免同时占用 AX Engine 资源。Web 服务以单个 AX Engine worker 运行，确保模型生命周期和并发请求安全。
性能统计
模型预加载已单独执行并排除；未计入 HTTP 上传、任务排队、浏览器轮询和 LLM 总结，音频测试长度：时长 1955.236 s（32:35）。
| 后端 | 端到端耗时 | 端到端 RTF | ASR 耗时 | ASR RTF（按全音频） |
|---|---|---|---|---|
| SenseVoice 模式 | 206.063 s（3:26） | 0.1054 | 101.090 s | 0.0517 |
| FireRed AED + Punc 模式 | 428.629 s（7:09） | 0.2192 | 332.480 s | 0.1700 |
RTF 越低越快。两者都满足离线快于实时：
- SenseVoice 端到端约 9.49× 实时；
- FireRed/Punc 端到端约 4.56× 实时。
(不同的音频长度和说话人数量会影响 RTF，以上仅为参考。)
运行时目录说明
AX Meeting Studio/
├── wheel/ # 为 AArch64 / 对应 CPython 版本构建的 ax_meeting_studio wheel
├── models/ # 外置推理模型：VAD、CAM++、SenseVoice、FireRed、Punc
├── certs/ # HTTPS 证书和私钥
├── src/ # 运行时资产清单等随部署保留的资源
├── tools/ # 模型资产完整性检查工具
├── .runtime/ # 首次启动后记录 wheel 安装状态（自动生成）
└── run_web_studio.sh # 预检、依赖/wheel 安装与 Web 服务启动入口
1. 下载
将整个
AX Meeting Studio/
目录下载到板端，例如：
cd /root
hf download AXERA-TECH/AX-Metting-Studio --local-dir AX-Metting-Studio
cd AX-Metting-Studio
板端需要具备 AArch64 AX8850N、CPython 3.10或CPython 3.13，以及
/soc/lib
下的 AX Engine 运行库。当前
包已带有证书；若部署环境不同，请替换
certs/cert.pem
和
certs/key.pem
。
cd AX-Metting-Studio
chmod 600 certs/key.pem
2. 启动
大模型总结功能需要在板端启动大模型后端服务，参考 AXERA-TECH/Qwen3.5-2B-AX650-GPTQ-Int4 进行下载和部署
首次启动会按需安装 wheel 与 Python 运行依赖。直接使用板端已有 BSP 运行库：
cd AX-Metting-Studio
AXMEETING_BSP_MSP_DIR=/soc ./run_web_studio.sh --port 8443
# output
AX Meeting Studio
URL: https://10.126.102.188:8443
Bind: 0.0.0.0:8443 (single AX Engine worker)
Backends: SenseVoice + FireRed AED/Punc
Models: /root/AX-Metting-Studio/models
Wheel: /root/AX-Metting-Studio/wheel/ax_meeting_studio-0.2.0-cp313-cp313-linux_aarch64.whl
Logs: uvicorn=warning, access=OFF, ax=WARN
Web UI: https://10.126.102.188:8443
浏览器访问Web UI：
https://<板端IP>:8443
--port
可改为任意未占用的
1-65535
端口；也可使用
PORT=8443
环境变量。
工作台界面：
离线会议转录和总结：
实时会议转录和总结：
3. 启动前检查
AXMEETING_BSP_MSP_DIR=/soc ./run_web_studio.sh --port 8443 --check
--check
不会安装依赖或 wheel，因此首次部署请先正常启动一次。
4. 常用配置
模型默认位于本目录的
models/
。模型在外接磁盘时：
AXMEETING_BSP_MSP_DIR=/soc \
AX_MODEL_DIR=/mnt/models \
./run_web_studio.sh --port 8443
自定义证书时：
AXMEETING_BSP_MSP_DIR=/soc \
SSL_CERT=/etc/ssl/cert.pem \
SSL_KEY=/etc/ssl/key.pem \
./run_web_studio.sh --port 8443
Python API 与 CLI 使用新命名：
from ax_meeting_studio import ModelBundle, MeetingEngine
ax-meeting-studio --help
Web UI 的“参数设置”可配置本地 OpenAI API、总结模型、模板及自定义要求。
TODO:
- AX8850算力卡支持
- 更多后端ASR模型
Runs of AXERA-TECH AX-Meeting-Studio on huggingface.co
31
Total runs
0
24-hour runs
-4
3-day runs
-4
7-day runs
-1
30-day runs
More Information About AX-Meeting-Studio huggingface.co Model
Url of AX-Meeting-Studio
AX-Meeting-Studio huggingface.co Url
Provider of AX-Meeting-Studio huggingface.co
Other API from AXERA-TECH
Updated:August 26 2026
Total runs:
214
Run Growth:
-116
Growth Rate:
-54.21%
Total runs:
207
Run Growth:
61
Growth Rate:
29.47%
Total runs:
152
Run Growth:
119
Growth Rate:
78.29%
Total runs:
91
Run Growth:
91
Growth Rate:
100.00%
Total runs:
91
Run Growth:
-389
Growth Rate:
-427.47%
Total runs:
61
Run Growth:
40
Growth Rate:
65.57%
Total runs:
60
Run Growth:
22
Growth Rate:
36.67%
Total runs:
58
Run Growth:
58
Growth Rate:
100.00%
Total runs:
46
Run Growth:
23
Growth Rate:
50.00%
Total runs:
44
Run Growth:
14
Growth Rate:
31.82%
Total runs:
44
Run Growth:
22
Growth Rate:
50.00%
Total runs:
42
Run Growth:
23
Growth Rate:
54.76%
Total runs:
42
Run Growth:
-4
Growth Rate:
-9.52%
Total runs:
42
Run Growth:
-3
Growth Rate:
-7.14%
Total runs:
42
Run Growth:
0
Growth Rate:
0.00%
Total runs:
42
Run Growth:
-198
Growth Rate:
-471.43%
Total runs:
39
Run Growth:
0
Growth Rate:
0.00%
Total runs:
39
Run Growth:
22
Growth Rate:
56.41%
Total runs:
38
Run Growth:
4
Growth Rate:
10.53%
Total runs:
35
Run Growth:
4
Growth Rate:
11.43%
Total runs:
35
Run Growth:
-1
Growth Rate:
-2.86%
Total runs:
34
Run Growth:
-10
Growth Rate:
-29.41%
Total runs:
34
Run Growth:
6
Growth Rate:
17.65%
Total runs:
34
Run Growth:
7
Growth Rate:
20.59%
Total runs:
33
Run Growth:
22
Growth Rate:
66.67%
Total runs:
33
Run Growth:
12
Growth Rate:
36.36%
Total runs:
33
Run Growth:
-6
Growth Rate:
-18.18%
Total runs:
33
Run Growth:
1
Growth Rate:
3.03%
Total runs:
32
Run Growth:
-4
Growth Rate:
-12.50%
Total runs:
32
Run Growth:
10
Growth Rate:
31.25%
Total runs:
32
Run Growth:
5
Growth Rate:
15.63%
Total runs:
31
Run Growth:
31
Growth Rate:
100.00%
Total runs:
31
Run Growth:
10
Growth Rate:
32.26%
Total runs:
31
Run Growth:
-20
Growth Rate:
-64.52%
Total runs:
30
Run Growth:
7
Growth Rate:
23.33%
Total runs:
30
Run Growth:
6
Growth Rate:
20.00%
Total runs:
29
Run Growth:
-4
Growth Rate:
-13.79%
Total runs:
29
Run Growth:
14
Growth Rate:
48.28%
Total runs:
29
Run Growth:
5
Growth Rate:
17.24%
Total runs:
29
Run Growth:
8
Growth Rate:
27.59%
Total runs:
28
Run Growth:
12
Growth Rate:
42.86%
Total runs:
28
Run Growth:
16
Growth Rate:
57.14%
Total runs:
28
Run Growth:
9
Growth Rate:
32.14%
Total runs:
27
Run Growth:
-47
Growth Rate:
-174.07%
Total runs:
27
Run Growth:
13
Growth Rate:
48.15%
Total runs:
27
Run Growth:
17
Growth Rate:
62.96%
Total runs:
27
Run Growth:
5
Growth Rate:
18.52%
Total runs:
26
Run Growth:
-20
Growth Rate:
-74.07%
Total runs:
25
Run Growth:
2
Growth Rate:
8.00%
Total runs:
23
Run Growth:
3
Growth Rate:
13.04%
Total runs:
22
Run Growth:
11
Growth Rate:
50.00%
Total runs:
22
Run Growth:
-10
Growth Rate:
-45.45%
Total runs:
21
Run Growth:
5
Growth Rate:
23.81%