SpeechColab ASR leaderboard

March 29, 2025 · View on GitHub

result result

1. Overview

"If you can’t measure it, you can’t improve it." -- Peter Drucker

SpeechIO leaderboard serves as an ASR benchmarking platform by providing 3 components:

  1. TestSet Zoo: A collection of test sets covering wide range of speech recognition tasks & scenarios

  2. Model Zoo: A collection of models including commercial APIs & open-sourced models

  3. Benchmarking Pipeline: a simple & well-specified pipeline to take care of data preparation / recognition / post processing / error rate evaluation.

People should be able to easily benchmark, reproduce, examine ASR systems from each other

Overview


2. TestSet Zoo: datasets/*

Academic Test Sets (EN & ZH)

已公开
UNLOCKED
编号
DATASET_ID
说明
DESCRIPTION
语言
LANGUAGE
AISHELL1_TESTtest set of AISHELL-1zh
AISHELL2_IOS_TESTtest set of AISHELL-2 (iOS channel)zh
AISHELL2_ANDROID_TESTtest set of AISHELL-2 (Android channel)zh
AISHELL2_MIC_TESTtest set of AISHELL-2 (Microphone channel)zh
ALIMEETING_EVAL_NEAR_FIELDAliMeetingzh
ALIMEETING_TEST_NEAR_FIELDAliMeetingzh
ALIMEETING_EVAL_FAR_FIELDAliMeetingzh
ALIMEETING_TEST_FAR_FIELDAliMeetingzh
LIBRISPEECH_TEST_CLEAN"test_clean" set of LibriSpeechen
LIBRISPEECH_TEST_OTHER"test_other" set of LibriSpeechen
TEDLIUM_RELEASE3_LEGACY_DEVtedlium release 3, legacy dir dev set TEDLium3en
TEDLIUM_RELEASE3_LEGACY_TESTtedlium release 3, legacy dir test set TEDLium3en
GIGASPEECH_V1.0.0_DEVdev set of GigaSpeechen
GIGASPEECH_V1.0.0_TESTtest set of GigaSpeechen
VOXPOPULI_V1.0_EN_DEVdev set of VoxPopulien
VOXPOPULI_V1.0_EN_TESTtest set of VoxPopulien
VOXPOPULI_V1.0_EN_ACCENTED_TESTaccented test set of VoxPopulien
COMMON_VOICE_V11.0_DEVdev set of Common Voiceen
COMMON_VOICE_V11.0_TESTtest set of Common Voiceen

SpeechIO Test Sets (ZH)

SpeechIO test sets are carefully curated by SpeechIO authors, crawled from publicly available sources (Youtube, TV programs, Podcast etc), covering various well-known scenarios and topics, transcribed by payed professional annotators.
已公开
UNLOCKED
编号
DATASET_ID
名称
NAME
场景
SCENARIO
内容领域
TOPIC
有效时长
DURATION (HOURS)
难度(1-5)
DIFFICULTY
SPEECHIO_ASR_ZH00000调试集
for debugging
视频会议、论坛演讲
conference & speech
经济、货币、金融
economy, currency, finance
1.0★★☆
SPEECHIO_ASR_ZH00001新闻联播新闻播报
TV News
时政
news & politics
9
SPEECHIO_ASR_ZH00002鲁豫有约访谈电视节目
TV interview
名人工作/生活
celebrity & film & music & daily
3★★☆
SPEECHIO_ASR_ZH00003天下足球专题电视节目
TV program
足球
Sports & Football & Worldcup
2.7★★☆
SPEECHIO_ASR_ZH00004罗振宇跨年演讲会场演讲
Stadium Public Speech
社会、人文、商业
Society & Culture & Business Trend
2.7★★
SPEECHIO_ASR_ZH00005李永乐讲堂在线教育
Online Education
科普
Popular Science
4.4★★★
SPEECHIO_ASR_ZH00006王者荣耀
张大仙 & 骚白
直播
Live Broadcasting
游戏
Game
1.6★★★☆
SPEECHIO_ASR_ZH00007直播带货
李佳琪 & 薇娅
直播
Live Broadcasting
电商、美妆
Makeup & Online shopping/advertising
0.9★★★★☆
SPEECHIO_ASR_ZH00008老罗语录线下培训
Offline lecture
段子、做人
Life & Purpose & Ethics
1.3★★★★☆
SPEECHIO_ASR_ZH00009故事FM播客
Podcast
人生故事、见闻
Ordinary Life Story Telling
4.5★★☆
SPEECHIO_ASR_ZH00010创业内幕播客
Podcast
创业、产品、投资
Startup & Enterprenuer & Product & Investment
4.2★★☆
SPEECHIO_ASR_ZH00011罗翔刑法法考在线教育
Online Education
法律 法考
Law & Lawyer Qualification Exams
3.4★★☆
SPEECHIO_ASR_ZH00012张雪峰考研在线教育
Online Education
考研 高校报考
University & Graduate School Entrance Exams
3.4★★★☆
SPEECHIO_ASR_ZH00013谷阿莫
牛叔说电影
短视频
VLog
电影剪辑
Movie Cuts
1.8★★★
SPEECHIO_ASR_ZH00014贫穷料理
琼斯爱生活
短视频
VLog
美食、烹饪
Food & Cooking & Gourmet
1★★★☆
SPEECHIO_ASR_ZH00015单田芳 白眉大侠评书
Traditional Podcast
江湖、武侠
Kongfu Fiction
2.2★★☆
SPEECHIO_ASR_ZH00016德云社演出剧场相声
Theater Crosstalk Show
包袱段子
Funny Stories
1★★★
SPEECHIO_ASR_ZH00017吐槽大会脱口秀电视节目
Standup Comedy
明星糗事
Celebrity Jokes
1.8★★☆
SPEECHIO_ASR_ZH00018小猪佩奇
熊出没
少儿动画
Children Cartoon
童话故事、日常
Fairy Tale
0.9★☆
SPEECHIO_ASR_ZH00019CCTV5 NBA 转播体育赛事解说
Sports Game Live
篮球、NBA
NBA Game
0.7★★★
SPEECHIO_ASR_ZH00020篮球人物纪录片
Documentary
篮球明星、成长
NBA Super Stars' Life & History
2.2★★
SPEECHIO_ASR_ZH00021汽车之家评测短视频
VLog
汽车测评
Car benchmarks, Road driving test
1.7★★★☆
SPEECHIO_ASR_ZH00022小艾大叔 豪宅带看短视频
VLog
房地产、豪宅
Realestate, Mansion tour
1.7★★★
SPEECHIO_ASR_ZH00023无聊开箱
Zealer评测
短视频
VLog
产品开箱评测
Unboxing
2★★★
SPEECHIO_ASR_ZH00024付老师种植技术短视频
VLog
农业、种植
Agriculture, Planting
2.7★★★☆
SPEECHIO_ASR_ZH00025石国鹏讲历史线下培训
Offline lecture
历史,古希腊哲学
History, Greek philosophy
1.3★★☆
SPEECHIO_ASR_ZH00026张震鬼故事广播节目
Broadcasting Program
鬼故事
Horror Stories
2.4★★★
SPEECHIO_ASR_ZH00027华语辩论世界杯辩论赛
Debates Contest
兴趣、技能、成长
Hobby, Skill, Growth
1.4★★★
SPEECHIO_ASR_ZH00028时政现场同传同声传译
Simultaneous Translation
时政、社会公共治理
News & Events on Public Governance
2.1★★★☆
SPEECHIO_ASR_ZH00029港台明星访谈
周杰伦,曾志伟
张家辉,陈小春
周星驰
口音(港台)
HongKong/Taiwan Accents
娱乐、生活、演艺
Entertainment, Acting, Musics
1.5★★★☆
SPEECHIO_ASR_ZH00030世界青年说口音(老外)
Foreigner Accents
异国文化比较
Cultural Difference
2★★★☆
SPEECHIO_ASR_ZH00031东方甄选直播
broadcast
带货,英语教学
Online advertising & English Education
2.4★★★☆
SPEECHIO_ASR_ZH00032郎朗钢琴课长视频
long-form video
音乐乐理,钢琴
Music & piano
1.7★★☆
SPEECHIO_ASR_ZH00033老石谈芯短视频
VLog
芯片
chips
2.8★★★
SPEECHIO_ASR_ZH00034电丸科技AK短视频
VLog
网络 IT
Internet tech, IT
1.4★★★☆
SPEECHIO_ASR_ZH00035新氧医美短视频
VLog
医疗美容
Medical Cosmetology
1.4★★
SPEECHIO_ASR_ZH00036交通广播交通广播
traffic radio
路况,娱乐
Traffics
1.2★★★☆
SPEECHIO_ASR_ZH00037老俞闲聊在线会议
Online meeting
闲聊
chat
2.4★★★
SPEECHIO_ASR_ZH00038电影:疯狂石头+疯狂赛车电影
Film
重庆话、山东青岛、四川成都话、河北唐山话、粤语、天津话、河南话、陕西话、闽南话,武汉话等
multiple accents
1.3★★★★☆
SPEECHIO_ASR_ZH00039电影:1942电影
Film
河南话
HeNan Accent
0.9★★★★
SPEECHIO_ASR_ZH00040电影:白鹿原电影
Film
陕西话
ShaanXi Accent
1.1★★★★★
SPEECHIO_ASR_ZH00041电影:让子弹飞电影
Film
四川话
SiChuan Accent
1.1★★★★☆
SPEECHIO_ASR_ZH00042电影:人生大事电影
Film
武汉话
WuHan Accent
0.8★★★★
SPEECHIO_ASR_ZH00043听障听障语音识别
Hearing Imperiment Speaker
新闻脚本
News Prompts
0.6★★★★★
SPEECHIO_ASR_ZH00044唐诗宋词诗词朗诵
Poems Reading
唐诗宋词
Chinese Poems
1.1★★★☆
SPEECHIO_ASR_ZH00045文言文文言文朗诵
Classical Chinese Reading
论语,老子,诗经,孙子兵法0.5★★★★★
SPEECHIO_ASR_ZH00046音乐歌词识别演唱
Singing
歌词
Lyrics
1.2★★★★☆

Download Dataset


3. Model Zoo: models/*

EN Models

编号
MODEL_ID
类型
TYPE
厂商/作者
PROVIDER/AUTHOR
简介
DESCRIPTION
链接
URL
aliyun_api_enCloudAlibabalink
amazon_api_enCloudAmazon AWSlink
baidu_api_enCloudBaidulink
google_api_enCloudGooglelink
google_USM_enCloudGooglerequest access
microsoft_sdk_enCloudMicrosoft Azurelink
tencent_api_enCloudTencentlink
coqui_model_enLocal
coquilink
deepspeech_model_enLocal
deepspeechlink
k2_gigaspeechLocal
k2-fsalink
nemo_conformer_ctc_large_enLocal
NVidia NeMolink
nemo_conformer_transducer_xlarge_enLocal
NVidia NeMolink
vosk_model_enLocal
alphacepheilink
vosk_model_en_largeLocal
alphacepheilink
whisper_largeLocal
OpenAIlink
whisper_large_v2Local
OpenAIlink
data2vec_audio_large_ft_libri_960hLocalFacebook AIlink
hubert_xlarge_ft_libri_960hLocalFacebook AIlink
wav2vec2_large_robust_ft_libri_960hLocalFacebook AIlink
wavlm_base_plus_ft_libri_clean_100hLocalMicrosoft
patrickvonplaten
link

ZH Models

Cloud Models

编号
MODEL_ID
类型
TYPE
厂商
PROVIDER
简介
DESCRIPTION
链接
URL
aispeech_api_zhCloud思必驰
AISpeech
思必驰开放平台link
aliyun_api_zhCloud阿里巴巴
Alibaba
阿里云 - 一句话识别link
aliyun_ftasr_api_zhCloud阿里巴巴
Alibaba
阿里云 - 文件识别(非流式)link
baidu_pro_api_zhCloud百度
Baidu
百度智能云
(极速版)
link
bilibili_api_zhCloud哔哩哔哩
bilibili
哔哩哔哩AI开放平台not available yet
ximalaya_api_zhCloud喜马拉雅
ximalaya
喜马拉雅AI开放平台
(转写,非流式)
link
iflytek_lfasr_api_zhCloud讯飞
IFlyTek
讯飞开放平台
(转写,非流式)
link
microsoft_sdk_zhCloud微软
Microsoft
Azure
(流式)
link
microsoft_batch_zhCloud微软
Microsoft
Azure
(离线转写)
link
tencent_api_zhCloud腾讯
Tencent
腾讯云link
yitu_api_zhCloud依图
YituTech
依图语音开放平台link

Local Models

编号
MODEL_ID
类型
TYPE
作者
AUTHOR
简介
DESCRIPTION
speechio_kaldi_multicnLocalXingyu NA(那兴宇)Kaldi multi_cn recipe
vosk_model_cnLocalalphacepheiChinese engine of Vosk
paraformer_large_offline_zhLocalmodelscopeParaformer, default Chinese 16k model, offline, support long-form audio recognition

Download Model

To submit a model

Follow this specification. Existing models are good references as well.


4. Benchmarking Pipeline

Benchmark


5. Latest Results

Public Models

Unlocked SpeechIO test sets (ZH00001 ~ ZH00026)

Rank 排名Model 模型CER 字错误率Date 时间
1ximalaya_api_zh1.72%2025.01
2aliyun_ftasr_api_zh1.80%2025.01
3microsoft_batch_zh1.95%2025.01
4iflytek_lfasr_api_zh3.01%2025.01
5tencent_api_zh3.20%2025.01
6aispeech_api_zh3.61%2025.01
7baidu_pro_api_zh7.30%2025.01

Locked SpeechIO test sets (ZH00027 ~ ZH00046)

Rank 排名Model 模型CER 字错误率Date 时间
1microsoft_batch_zh5.26%2025.01
2ximalaya_api_zh6.89%2025.01
3aliyun_ftasr_api_zh6.92%2025.01
4tencent_api_zh7.81%2025.01
5iflytek_lfasr_api_zh8.70%2025.01
6aispeech_api_zh10.42%2025.01
7baidu_pro_api_zh16.23%2025.01

All SpeechIO test sets (ZH00001 ~ ZH00046)

Rank 排名Model 模型CER 字错误率Date 时间
1microsoft_batch_zh2.99%2025.01
2ximalaya_api_zh3.35%2025.01
3aliyun_ftasr_api_zh3.40%2025.01
4tencent_api_zh4.64%2025.01
5iflytek_lfasr_api_zh4.80%2025.01
6aispeech_api_zh5.75%2025.01
7baidu_pro_api_zh10.10%2025.01

Private Models

Unlocked SpeechIO test sets (ZH00001 ~ ZH00026)

Model 模型CER 字错误率Date 时间
bilibili_api_zh(*)2.49%2025.01

Locked SpeechIO test sets (ZH00027 ~ ZH00046)

Model 模型CER 字错误率Date 时间
bilibili_api_zh(*)5.56%2025.01

All SpeechIO test sets (ZH00001 ~ ZH00046)

Model 模型CER 字错误率Date 时间
bilibili_api_zh(*)3.45%2025.01
Detail all results (字错误率 CER %)

Test Set ID测试场景&内容领域bilibili_api_zhDate 时间
SPEECHIO_ASR_ZH00001新闻联播0.532025.01
SPEECHIO_ASR_ZH00002访谈2.832025.01
SPEECHIO_ASR_ZH00003电视节目0.972025.01
SPEECHIO_ASR_ZH00004场馆演讲1.592025.01
SPEECHIO_ASR_ZH00005在线教育1.452025.01
SPEECHIO_ASR_ZH00006直播5.762025.01
SPEECHIO_ASR_ZH00007直播6.402025.01
SPEECHIO_ASR_ZH00008线下培训3.692025.01
SPEECHIO_ASR_ZH00009播客3.182025.01
SPEECHIO_ASR_ZH00010播客3.482025.01
SPEECHIO_ASR_ZH00011在线教育1.782025.01
SPEECHIO_ASR_ZH00012在线教育2.132025.01
SPEECHIO_ASR_ZH00013短视频3.032025.01
SPEECHIO_ASR_ZH00014短视频3.472025.01
SPEECHIO_ASR_ZH00015评书4.832025.01
SPEECHIO_ASR_ZH00016相声3.042025.01
SPEECHIO_ASR_ZH00017脱口秀2.822025.01
SPEECHIO_ASR_ZH00018少儿卡通1.962025.01
SPEECHIO_ASR_ZH00019体育赛事解说2.292025.01
SPEECHIO_ASR_ZH00020纪录片1.552025.01
SPEECHIO_ASR_ZH00021短视频1.692025.01
SPEECHIO_ASR_ZH00022短视频3.472025.01
SPEECHIO_ASR_ZH00023短视频2.142025.01
SPEECHIO_ASR_ZH00024短视频4.702025.01
SPEECHIO_ASR_ZH00025线下课堂3.142025.01
SPEECHIO_ASR_ZH00026广播电台节目3.632025.01
SPEECHIO_ASR_ZH00027华语大学生辩论赛2.032025.01
SPEECHIO_ASR_ZH00028同声传译:时政&社会公共治理2.042025.01
SPEECHIO_ASR_ZH00029港台口音:港台明星访谈3.872025.01
SPEECHIO_ASR_ZH00030老外口音:《世界青年说》3.862025.01
SPEECHIO_ASR_ZH00031直播带货3.742025.01
SPEECHIO_ASR_ZH00032音乐3.862025.01
SPEECHIO_ASR_ZH00033芯片2.452025.01
SPEECHIO_ASR_ZH00034网络IT5.102025.01
SPEECHIO_ASR_ZH00035新氧医美1.132025.01
SPEECHIO_ASR_ZH00036交通广播6.012025.01
SPEECHIO_ASR_ZH00037在线会议聊天3.022025.01
SPEECHIO_ASR_ZH00038电影:疯狂石头+疯狂赛车(方言杂烩)18.362025.01
SPEECHIO_ASR_ZH00039电影:1942(河南话)13.922025.01
SPEECHIO_ASR_ZH00040电影:白鹿原(陕西话)25.802025.01
SPEECHIO_ASR_ZH00041电影:让子弹飞(四川话)11.372025.01
SPEECHIO_ASR_ZH00042电影:人生大事(武汉话)18.242025.01
SPEECHIO_ASR_ZH00043听障23.342025.01
SPEECHIO_ASR_ZH00044诗词1.642025.01
SPEECHIO_ASR_ZH00045文言文4.222025.01
SPEECHIO_ASR_ZH00046歌词9.602025.01

note: models with (*) marker can be found in model zoo, but not universally available to public yet.


Contacts

Email: leaderboard@speechio.ai