Instructions to use AutoArk-AI/ARK-ASR-3B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use AutoArk-AI/ARK-ASR-3B with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("automatic-speech-recognition", model="AutoArk-AI/ARK-ASR-3B", trust_remote_code=True)# Load model directly from transformers import AutoModelForCausalLM model = AutoModelForCausalLM.from_pretrained("AutoArk-AI/ARK-ASR-3B", trust_remote_code=True, dtype="auto") - Notebooks
- Google Colab
- Kaggle
Will you upload the ARK-ASR-3B (Transformers native) version?
There is an issue with the trust_remote_code=True path!The expected correct outcome was to transcribe and generate Japanese subtitles, but the final result was garbled text! It involved at least 15 different languages!It's exactly the same issue I ran into while using another ASR model!https://huggingface.co/CohereLabs/cohere-transcribe-03-2026/discussions/28
Loading ARK-ASR-3B model from C:\Users\supervip.cache\huggingface\hub\models--AutoArk-AI--ARK-ASR-3B\snapshots\1e28271b79edc97635783bea65abc89195a09ed3...
Loading weights: 100%|█████████████████████████████████████████████████████████████| 926/926 [00:00<00:00, 3131.44it/s]
1
00:00:15,500 --> 00:00:16,199
三万块的
2
00:00:49,000 --> 00:00:50,899
あの時のお礼が言いたいだけ
3
00:01:12,700 --> 00:01:17,074
ano fujima
san nara kou
4
00:01:17,074 --> 00:01:18,299
chou
desu kedo ah
5
00:01:18,900 --> 00:01:21,000
お疲れ様で
す 総務の綾川です
6
00:01:21,799 --> 00:01:23,400
新しい名所を私に来たんで
7
00:01:25,500 --> 00:01:26,900
わかりましょうか.
8
00:01:27,099 --> 00:01:28,472
直接渡すのが決まりなのでで
9
00:01:28,472 --> 00:01:29,844
も皆さん置いてきますよめん
10
00:01:29,844 --> 00:01:32,799
どくさいでしょう大丈夫です
11
00:01:33,000 --> 00:01:34,099
また明日きます
12
00:01:34,599 --> 00:01:36,700
ご親切にあ
りがとうございます
13
00:01:51,200 --> 00:01:52,000
sasajima
shi ni
14
00:02:06,400 --> 00:02:12,000
sumimasen.
新しい名刺を渡しに来ました
15
00:02:12,199 --> 00:02:14,599
こちらに受け取
りのサインをお願いします
16
00:02:15,199 --> 00:02:17,099
バスカルキラしてたから
17
00:02:17,300 --> 00:02:22,400
thank you
very much. 是的
18
00:02:24,000 --> 00:02:24,500
hi.
19
00:02:53,699 --> 00:02:54,845
was a new
employee, a
20
00:02:54,845 --> 00:03:00,000
kawa aiko. so
21
00:03:04,699 --> 00:03:05,900
私の恩人です
22
00:03:06,300 --> 00:03:08,900
と俺をいたか
ったんです 洪总,你
23
00:03:09,599 --> 00:03:11,800
いやいや俺は別に
24
00:03:12,599 --> 00:03:17,400
我们都在啊 you c
an. yeah.
25
00:03:47,500 --> 00:03:50,500
し見つけち
ゃった nanyo
26
00:03:51,599 --> 00:03:55,900
この前話したでし
ょ正一の結婚相手 Oh.
27
00:03:56,400 --> 00:03:57,400
hong kong
28
00:03:59,699 --> 00:04:03,900
three
more i see
29
00:04:04,599 --> 00:04:07,900
特されなんか分かれてくれそ
うなこう見つけたから
30
00:04:11,599 --> 00:04:18,000
この 男性経験がなさそうで
まじめで簡単に落ちそう
31
00:04:20,000 --> 00:04:26,500
私の条件を完璧に
クリアしてるんだよね 都
32
00:04:33,199 --> 00:04:34,100
jimi suki
33
00:04:35,899 --> 00:04:38,000
他にも配偶者はいるんだけど
34
00:04:38,300 --> 00:04:39,500
送过他的生日
35
00:04:43,500 --> 00:04:44,935
トウキノクの話によると家族
36
00:04:44,935 --> 00:04:46,370
の会話とかで飲み会にはほと
37
00:04:46,370 --> 00:04:48,800
んど参加しないみたい
38
00:04:59,100 --> 00:05:01,100
でも道場なんてしないでよね
39
00:05:02,399 --> 00:05:04,500
これはただの
カムフラージュなんだろ
40
00:05:08,800 --> 00:05:09,300
what do you
41
00:05:19,399 --> 00:05:21,199
本気で好きになるわけないか
42
00:05:23,800 --> 00:05:27,100
itadakimasu.
I got some.
43
00:05:28,199 --> 00:05:29,399
お疲れ様です
44
00:05:33,000 --> 00:05:34,800
yeah
45
00:05:35,000 --> 00:05:37,600
大したものじゃないです サ
ボるときもありますし
46
00:05:42,500 --> 00:05:46,000
yeah.
as you saw
47
00:05:48,899 --> 00:05:51,100
昼飯一緒にどうかな
48
00:05:51,399 --> 00:05:54,800
仕事の話もゆ
っくりしたいし 让江
49
00:05:55,500 --> 00:05:56,399
what is it
50
00:05:58,600 --> 00:05:59,600
むむりしないだ
51
00:06:00,399 --> 00:06:01,995
部署が違ってもね一応上司と
52
00:06:01,995 --> 00:06:03,590
部下だしハラスメントみたい
53
00:06:03,590 --> 00:06:05,800
になっても困るし
54
00:06:08,800 --> 00:06:14,199
thank you.
i don't know
55
00:06:15,000 --> 00:06:15,500
or the
Hi, we are working on the transformers integration PR ;). Please stay tuned.