實際影片長度:7:04.000。原文、繁中、雙語可點擊句子跳轉影片。
0:00.000–0:05.480
This little dictation app is free, open source, and
0:02.740–0:05.480
uses AI on your computer.
0:05.680–0:08.740
And I'm going to show you how you can install it and use it right now.
0:09.360–0:12.360
Presented by Zapier, the AI automation company.
0:12.920–0:15.660
Adam, everyone's using audio dictation apps on their computers.
0:15.780–0:17.440
They're paying 15 bucks a month for it.
0:17.500–0:20.180
I'm using Whisperflow, and we've got a better solution,
0:20.320–0:22.380
one that's completely free and works way better.
0:22.660–0:23.740
What do you use? What do you pay?
0:24.260–0:26.300
I use this thing called Super Whisper.
0:26.460–0:28.980
It's using all the same models that all these other tools use.
0:28.980–0:34.580
And I'm paying, it's about $10 a month, but
0:31.780–0:34.580
I don't like to think about it too much.
0:34.960–0:36.160
10 is fine.
0:36.260–0:39.840
I pay 15 bucks a month because I pay monthly for this.
0:40.000–0:41.760
I don't like it for a few reasons.
0:41.900–0:43.780
Number one, I'm sending it up into the cloud,
0:43.900–0:46.040
which I don't have any worries about any private information,
0:46.120–0:48.720
but who knows who's going to subpoena what that I said, number one.
0:48.940–0:51.960
Number two, if it's going up into the cloud, it's taking longer.
0:52.280–0:55.360
And number three,
0:53.820–0:55.360
I happen to believe that there's a better user experience,
0:55.440–0:57.240
and Whisperflow is not giving it to us.
0:57.240–0:58.960
And neither are these other apps.
0:59.060–1:03.080
So you and I did the GitHub show, and
1:01.070–1:03.080
we talked about the top repos of the week.
1:03.160–1:07.880
And then I said,
1:04.733–1:06.307
I've got to atta
1:06.307–1:07.880
ch something to it that people are not starring enough,
1:08.340–1:09.400
and it's this.
1:09.560–1:11.060
It is something called Fluid Voice.
1:11.160–1:14.800
Look, it shot up after we did our GitHub show,
1:12.980–1:14.800
which is incredibly popular.
1:14.960–1:16.800
If you're not watching it, folks, you've got to watch it.
1:16.800–1:18.900
We reveal these types of things all the time.
1:19.120–1:20.700
I'm one of those stars, Andrew.
1:20.860–1:23.480
I'm that star right at the kind of top third there.
1:23.600–1:26.040
I was like, oh, man, after I tried it and I started using it,
1:26.240–1:29.640
I immediately came back to the repository, gave it a star,
1:29.740–1:32.060
because I think these guys are going to be really successful.
1:32.240–1:33.020
It's an awesome product.
1:33.140–1:37.200
It's fantastic because you get to customize it.
1:37.200–1:43.060
You get to download this or get the whole code base,
1:43.160–1:44.740
and you could change it to however you want.
1:44.840–1:46.980
Or you could just go to their website, and you can get it.
1:47.080–1:48.780
I got it and downloaded it for Mac.
1:49.040–1:51.100
The setup is super-duper easy.
1:51.580–1:54.820
Then once I had it, I got to pick which voice engine I wanted.
1:54.880–1:56.200
And I said, I want the blazing fast one.
1:56.280–2:00.040
You can go for whatever model you like.
2:00.040–2:05.560
And then you can customize it here by telling it what you want to use as your hot key,
2:05.740–2:08.740
what you want to use as your continuation key.
2:08.820–2:11.300
It will even do some edits on your text if you want it.
2:11.540–2:15.220
Let me show you what it looks like compared to Whisperflow.
2:16.000–2:18.120
On Whisperflow, I'm going to hit the FN key, and I'm going to say,
2:18.620–2:21.860
my name is Andrew, and
2:20.240–2:21.860
I'm just going to be talking directly into the screen,
2:21.980–2:24.180
and I'm going to talk super fast because I'm a New Yorker at heart,
2:24.260–2:25.400
and I will always talk super fast.
2:25.460–2:28.440
And you can see on the very bottom of my screen how all my WAV files are coming in.
2:28.440–2:29.880
Let's see how long it takes for it to come through.
2:30.800–2:31.180
Let go.
2:31.440–2:31.920
It's thinking.
2:32.420–2:33.360
Not bad, right?
2:34.240–2:34.420
Yeah.
2:34.680–2:34.840
Okay.
2:35.100–2:36.460
I think it basically got it.
2:36.680–2:40.340
Now I'm going to switch over to Fluid and watch this.
2:41.720–2:44.840
My name is Andrew, and
2:43.280–2:44.840
now I'm going to be using this Fluid voice thing.
2:44.920–2:49.580
And notice on the bottom of my screen,
2:47.250–2:49.580
you can not just see that it's typing with the WAV thing,
2:49.660–2:54.020
but it's also showing the words,
2:51.840–2:54.020
which lets me see that what I'm saying is actually going through.
2:54.240–2:58.840
And it's letting me see what kind of features I've got on and
2:56.540–2:58.840
where it's planning to paste it.
2:58.840–3:01.460
In this case, into Chrome.
3:01.840–3:05.260
Let's let go of the button, the hotkey, and
3:03.550–3:05.260
let's see how fast it pastes it in.
3:07.760–3:08.380
That's amazing.
3:08.440–3:08.660
Right.
3:09.120–3:11.240
And the fact that I can see it in real time is fantastic.
3:11.900–3:13.060
Really good piece of software.
3:13.060–3:15.920
I'm so glad that we got to share this with people.
3:16.420–3:17.720
What do you have to add on this?
3:17.760–3:19.280
You've been using this a lot longer than I have.
3:19.360–3:20.700
I mean, there's so much.
3:20.960–3:24.220
So number one, I think you glossed over something I think is really important here.
3:24.400–3:29.760
You know, Whisperflow, Super Whisper, all these other ones,
3:27.080–3:29.760
they're all companies that are charging us to use this software.
3:29.760–3:32.700
The models are either open or available.
3:33.400–3:36.440
And these companies have packaged it into a nice Mac app for you.
3:37.060–3:38.580
FluidVoice is open source.
3:38.940–3:40.000
You could download the source.
3:40.100–3:40.980
You could save it forever.
3:41.140–3:42.920
You don't ever have to pay these guys money.
3:43.260–3:46.160
Other people are going to come in here and contribute to it.
3:46.200–3:47.100
They're going to change it.
3:47.280–3:48.420
There's going to be versions of this.
3:48.420–3:52.900
The fact that this is open means that this is free and
3:50.660–3:52.900
it's going to be free.
3:53.240–3:54.760
And you get the control.
3:54.900–3:57.820
Now, you don't have to customize it if you don't want to, but you could.
3:58.340–4:12.840
But what you've already pointed out is when
4:01.240–4:04.140
you go into the settings, there's a lot of
4:04.140–4:07.040
customization available there, including, y
4:07.040–4:09.940
ou know, running all of your transcriptions
4:09.940–4:12.840
through AI prompts before they get pasted into the application.
4:13.020–4:17.580
And so you could have set up, I want a prompt for writing emails.
4:17.580–4:20.080
Well, now it's going to remove your ticks.
4:20.200–4:23.100
It's going to add, you know, a new line after Dear Andrew.
4:23.500–4:25.080
I love the show that we do together.
4:25.400–4:27.100
It'll add the new line there.
4:27.160–4:27.920
It'll add the comma.
4:28.540–4:30.440
And you can customize those prompts.
4:30.800–4:32.300
You can also customize the dictionary.
4:32.680–4:42.240
I've already, when I get one of these tools,
4:35.070–4:37.460
the first thing I do is I drop into the cust
4:37.460–4:39.850
om words and I add in all the names of the p
4:39.850–4:42.240
eople that I work with that I say out loud all the time.
4:42.300–4:45.600
I add the names of the companies that I work with because
4:43.950–4:45.600
they're all spelled weird.
4:45.600–4:53.980
And immediately, once you put them in this dictionary,
4:49.790–4:53.980
again, fluid voice is going to use AI to plug in those words.
4:54.380–5:00.700
And so it's never going to spell your name wrong when
4:57.540–5:00.700
I speak to it because it knows exactly how Andrew Warner is spelled.
5:01.000–5:02.020
It knows how my name is spelled.
5:02.120–5:04.400
My partner, Jesse Poogee, it knows all of that.
5:04.480–5:05.780
Is it J-E-S-S-I?
5:05.900–5:06.860
J-E-S-I-E?
5:07.200–5:07.980
J-E-S-S-E?
5:08.160–5:09.680
Like fluid voice knows.
5:09.680–5:11.940
So I've put all of that in place.
5:12.200–5:13.080
And I agree.
5:13.160–5:14.740
I'm on the Blazing Fast model, too.
5:14.900–5:17.700
But I did actually test out a couple of the other models.
5:17.880–5:20.460
I tested one of the even faster ones.
5:20.820–5:23.300
I also tested one that was, I don't remember which it was.
5:23.340–5:25.200
Yeah, the Cohere High Accuracy.
5:25.380–5:27.500
It got like two more accuracy points.
5:28.360–5:29.740
But it's a little slower.
5:30.200–5:31.860
And I decided, no, I can't tell a difference.
5:32.020–5:33.880
And so I'm on the same one that you're on.
5:34.420–5:39.140
The only weird thing, you know, when I first downloaded this,
5:36.780–5:39.140
I installed it.
5:39.300–5:40.600
I clicked Blazing Fast.
5:40.700–5:41.700
It started to install.
5:42.340–5:43.680
And then nothing happened.
5:43.860–5:44.680
And nothing happened.
5:44.880–5:46.540
And I actually wrote you a Slack message.
5:46.960–5:49.560
It was like, oh, man, Andrew, like, I was wrong.
5:49.740–5:51.880
In the video, I said this is going to be amazing.
5:52.540–5:53.460
And it's not.
5:53.540–5:54.060
I'm sorry.
5:54.220–5:54.880
But you know what?
5:54.900–5:56.200
I waited a couple more minutes.
5:56.200–5:57.520
The thing installed.
5:57.660–5:58.820
And it totally works.
5:58.940–6:02.060
And I've already removed Super Whisper from my computer.
6:02.060–6:03.180
That's a good note.
6:03.260–6:04.440
Because the same thing happened to me.
6:04.500–6:06.580
And because it happened to you, I knew to watch out for it.
6:06.600–6:10.580
What happens is it will say 3% installed.
6:10.760–6:11.980
And then it just stops.
6:12.300–6:13.240
And it won't do anything.
6:13.760–6:19.800
And then a few minutes later, maybe 10, 15 minutes later,
6:16.780–6:19.800
it shifts from 3% to 98 to 100.
6:20.020–6:20.860
And then it's done.
6:21.480–6:23.020
So that part wasn't working well.
6:23.020–6:28.220
But little inconveniences to me like that are going to be worth the $10 a month.
6:28.220–6:32.000
Plus, like, since it's installed and since it's running,
6:30.110–6:32.000
I've had zero issues.
6:32.460–6:33.360
Everything has worked great.
6:33.440–6:34.680
The little pop-up works great.
6:34.860–6:40.720
I love that you can see the text as you're speaking instead
6:37.790–6:40.720
of just the little thing that's moving.
6:40.820–6:43.320
Because half the time, like, wait, can it actually hear me?
6:43.540–6:46.280
Is it connected to my Bluetooth headphones that are in the other room?
6:46.540–6:48.100
Like, is this working?
6:48.460–6:51.200
Seeing the text on the screen is amazing, too.
6:51.440–6:52.100
And it's free.
6:52.320–6:53.360
Did we mention that it's free?
6:53.480–6:54.420
I love that it's free.
6:54.540–6:55.640
I love that it's open source.
6:55.640–6:59.200
Fluid Voice is one of many apps that we showed in the GitHub repo show.
6:59.500–7:01.240
Click the link here and we'll show you the others.
7:01.620–7:02.020
See you there.
0:00.000–0:05.480
這個小小的聽寫應用程式是免費且開源的,並且
0:02.740–0:05.480
在你的電腦上使用 AI。
0:05.680–0:08.740
我將向你展示如何立即安裝並使用它。
0:09.360–0:12.360
由 AI 自動化公司 Zapier 呈現。
0:12.920–0:15.660
亞當,現在大家都在電腦上使用音訊聽寫應用程式。
0:15.780–0:17.440
他們為此每月支付 15 美元。
0:17.500–0:20.180
我使用的是 Whisperflow,而我們有更好的解決方案,
0:20.320–0:22.380
一個完全免費且效能更好的方案。
0:22.660–0:23.740
你使用什麼?你支付多少?
0:24.260–0:26.300
我使用一個叫做 Super Whisper 的東西。
0:26.460–0:28.980
它使用了與所有其他工具相同的模型。
0:28.980–0:34.580
我每月支付大約 10 美元,但
0:31.780–0:34.580
我不喜歡想太多。
0:34.960–0:36.160
10 美元沒問題。
0:36.260–0:39.840
我每月支付 15 美元,因為我是按月付費。
0:40.000–0:41.760
我不喜歡它有幾個原因。
0:41.900–0:43.780
第一,我將它上傳到雲端,
0:43.900–0:46.040
雖然我對私人資訊沒有擔憂,
0:46.120–0:48.720
但誰知道誰會為了我說了什麼而發出法庭傳喚,這是第一點。
0:48.940–0:51.960
第二,如果它要上傳到雲端,速度會比較慢。
0:52.280–0:55.360
第三,
0:53.820–0:55.360
我認為有更好的使用者體驗,
0:55.440–0:57.240
而 Whisperflow 沒有提供給我們。
0:57.240–0:58.960
其他應用程式也沒有。
0:59.060–1:03.080
所以你和我們做了 GitHub 節目,並且
1:01.070–1:03.080
我們討論了每週的前幾名程式庫。
1:03.160–1:07.880
然後我說,
1:04.733–1:06.307
我必須
1:06.307–1:07.880
為它添加一些人們不夠關注的功能,
1:08.340–1:09.400
就是這個。
1:09.560–1:11.060
它叫做 Fluid Voice。
1:11.160–1:14.800
看,它在我们做了 GitHub 節目後迅速上升,
1:12.980–1:14.800
這非常受歡迎。
1:14.960–1:16.800
如果你還沒看,各位,你一定要看。
1:16.800–1:18.900
我們經常揭露這類事物。
1:19.120–1:20.700
我是那些星標之一,安德魯。
1:20.860–1:23.480
我就是那個位於頂部三分之一的星標。
1:23.600–1:26.040
我想說,喔,天啊,在我嘗試並開始使用它之後,
1:26.240–1:29.640
我立刻回到該程式庫,給了它一個星標,
1:29.740–1:32.060
因為我認為這些人會非常成功。
1:32.240–1:33.020
這是一款超讚的產品。
1:33.140–1:37.200
它很棒,因為你可以自訂它。
1:37.200–1:43.060
你可以下載這個或取得整個程式碼庫,
1:43.160–1:44.740
然後按照你的喜好進行修改。
1:44.840–1:46.980
或者你只需前往他們的網站,就可以取得它。
1:47.080–1:48.780
我下載並安裝了 Mac 版本。
1:49.040–1:51.100
設定過程超級簡單。
1:51.580–1:54.820
接著,當我安裝好之後,我必須選擇想要的語音引擎。
1:54.880–1:56.200
我說,我想要那個速度飛快的那個。
1:56.280–2:00.040
你可以選擇任何你喜歡的模型。
2:00.040–2:05.560
然後你可以在這裡進行自訂,告訴它你想使用什麼作為熱鍵,
2:05.740–2:08.740
你想使用什麼作為繼續鍵。
2:08.820–2:11.300
如果你希望,它甚至會對你的文字進行一些編輯。
2:11.540–2:15.220
讓我展示一下它與 Whisperflow 相比的樣子。
2:16.000–2:18.120
在 Whisperflow 上,我會按下 FN 鍵,然後我會說,
2:18.620–2:21.860
我的名字是安德魯,而且
2:20.240–2:21.860
我會直接對著螢幕說話,
2:21.980–2:24.180
我會說得非常快,因為我骨子里是個紐約人,
2:24.260–2:25.400
而且我總是說得非常快。
2:25.460–2:28.440
你可以在螢幕最下方看到我所有的 WAV 檔案是如何進來的。
2:28.440–2:29.880
讓我們看看它需要多長時間才能處理完成。
2:30.800–2:31.180
放開手。
2:31.440–2:31.920
它正在思考。
2:32.420–2:33.360
還不錯,對吧?
2:34.240–2:34.420
是的。
2:34.680–2:34.840
好的。
2:35.100–2:36.460
我想它基本上已經處理好了。
2:36.680–2:40.340
現在我要切換到 Fluid,請看。
2:41.720–2:44.840
我的名字是安德魯,而且
2:43.280–2:44.840
現在我要使用這個 Fluid 語音功能。
2:44.920–2:49.580
注意看我的螢幕底部,
2:47.250–2:49.580
你不僅可以看到它正在透過 WAV 進行輸入,
2:49.660–2:54.020
它還會顯示文字,
2:51.840–2:54.020
這讓我看到我說的內容確實正在被處理。
2:54.240–2:58.840
它讓我看到我啟用了哪些功能,
2:56.540–2:58.840
以及它計劃將文字貼到哪裡。
2:58.840–3:01.460
在這種情況下,是貼到 Chrome 瀏覽器。
3:01.840–3:05.260
讓我們放開按鈕,也就是熱鍵,然後
3:03.550–3:05.260
讓我們看看它貼上文字的速度有多快。
3:07.760–3:08.380
這太驚人了。
3:08.440–3:08.660
沒錯。
3:09.120–3:11.240
而且我能即時看到結果,這太棒了。
3:11.900–3:13.060
這套軟體真的很好用。
3:13.060–3:15.920
我很高興我們能與大家分享這個工具。
3:16.420–3:17.720
你還有什麼要補充的嗎?
3:17.760–3:19.280
你使用這套工具的經驗比我久多了。
3:19.360–3:20.700
我的意思是,這裡的功能太多了。
3:20.960–3:24.220
首先,我認為你略過了一個我認為非常重要的部分。
3:24.400–3:29.760
你知道的,Whisperflow、Super Whisper,還有其他這些軟體,
3:27.080–3:29.760
它們都是向我們收費以使用這些軟體的公司。
3:29.760–3:32.700
這些模型不是開源的,就是可以取得的。
3:33.400–3:36.440
這些公司將它們包裝成精美的 Mac 應用程式提供給你。
3:37.060–3:38.580
FluidVoice 是開源的。
3:38.940–3:40.000
你可以下載原始碼。
3:40.100–3:40.980
你可以永遠保存它。
3:41.140–3:42.920
你永遠不需要付錢給這些公司。
3:43.260–3:46.160
其他人會進來貢獻力量。
3:46.200–3:47.100
他們會修改它。
3:47.280–3:48.420
會有這個軟體的不同版本。
3:48.420–3:52.900
它是開源的這個事實意味著它是免費的,而且
3:50.660–3:52.900
它將永遠免費。
3:53.240–3:54.760
而且你擁有控制權。
3:54.900–3:57.820
現在,如果你不想自訂,你不需要自訂,但你可以。
3:58.340–4:12.840
但你已經指出的重點是,當
4:01.240–4:04.140
你進入設定時,那裡有很多
4:04.140–4:07.040
可用的自訂選項,包括,
4:07.040–4:09.940
你知道,在文字被貼上到應用程式之前,
4:09.940–4:12.840
先透過 AI 提示進行所有轉錄。
4:13.020–4:17.580
所以你可以設定,我想要一個撰寫電子郵件的提示。
4:17.580–4:20.080
現在,它會移除你的口頭語。
4:20.200–4:23.100
它會加入,你知道,在「親愛的 Andrew」之後換行。
4:23.500–4:25.080
我喜歡我們一起做的節目。
4:25.400–4:27.100
它會在這裡加入換行。
4:27.160–4:27.920
它會加入逗號。
4:28.540–4:30.440
你可以自訂這些提示。
4:30.800–4:32.300
你也可以自訂詞典。
4:32.680–4:42.240
我已經,當我拿到這些工具時,
4:35.070–4:37.460
我做的第一件事就是進入自訂
4:37.460–4:39.850
詞彙,並加入所有我
4:39.850–4:42.240
經常大聲說出的同事的名字。
4:42.300–4:45.600
我會加入我合作的公司名稱,因為
4:43.950–4:45.600
它們的拼法都很奇怪。
4:45.600–4:53.980
而且一旦你把它們放進這個字典,
4:49.790–4:53.980
Fluid Voice 就會再次使用 AI 來填入這些單字。
4:54.380–5:00.700
所以當我對它說話時,它永遠不會拼錯你的名字,因為
4:57.540–5:00.700
它非常清楚 Andrew Warner 的拼法。
5:01.000–5:02.020
它知道我的名字怎麼拼。
5:02.120–5:04.400
我的合夥人 Jesse Poogee,它都知道。
5:04.480–5:05.780
是 J-E-S-S-I 嗎?
5:05.900–5:06.860
還是 J-E-S-I-E?
5:07.200–5:07.980
或是 J-E-S-S-E?
5:08.160–5:09.680
Fluid Voice 都知道。
5:09.680–5:11.940
所以我把這些都設定好了。
5:12.200–5:13.080
我也同意。
5:13.160–5:14.740
我也在使用 Blazing Fast 模型。
5:14.900–5:17.700
但我確實測試過其他幾個模型。
5:17.880–5:20.460
我測試了一個更快的模型。
5:20.820–5:23.300
我也測試了一個……我記不起來是哪一個。
5:23.340–5:25.200
對,是 Cohere High Accuracy。
5:25.380–5:27.500
準確度只多了兩分。
5:28.360–5:29.740
但速度比較慢。
5:30.200–5:31.860
我決定,不,我感覺不出差別。
5:32.020–5:33.880
所以我跟你用同一個。
5:34.420–5:39.140
唯一奇怪的是,你知道,我第一次下載這個的時候,
5:36.780–5:39.140
我安裝了它。
5:39.300–5:40.600
我點擊了 Blazing Fast。
5:40.700–5:41.700
它開始安裝。
5:42.340–5:43.680
然後就沒反應了。
5:43.860–5:44.680
一直沒反應。
5:44.880–5:46.540
我甚至給你發了 Slack 訊息。
5:46.960–5:49.560
我說:「喔,天啊,Andrew,我想我錯了。」
5:49.740–5:51.880
在影片裡我說這會很棒。
5:52.540–5:53.460
但事實並非如此。
5:53.540–5:54.060
對不起。
5:54.220–5:54.880
但你知道嗎?
5:54.900–5:56.200
我又等了幾分鐘。
5:56.200–5:57.520
安裝完成了。
5:57.660–5:58.820
它完全正常運作。
5:58.940–6:02.060
我已經從電腦上移除了 Super Whisper。
6:02.060–6:03.180
這是個好提醒。
6:03.260–6:04.440
因為我也遇到了同樣的情況。
6:04.500–6:06.580
因為你也遇到了,所以我會特別留意這個問題。
6:06.600–6:10.580
情況是它會顯示已安裝 3%。
6:10.760–6:11.980
然後它就停住了。
6:12.300–6:13.240
什麼事都不會發生。
6:13.760–6:19.800
然後幾分鐘後,也許是 10 或 15 分鐘後,
6:16.780–6:19.800
它會從 3% 跳到 98% 再到 100%。
6:20.020–6:20.860
然後就完成了。
6:21.480–6:23.020
所以那部分運作得不太好。
6:23.020–6:28.220
但對我來說,像這樣的小不便,換取每月 10 美元的代價是值得的。
6:28.220–6:32.000
另外,既然它已經安裝並正在運行,
6:30.110–6:32.000
我完全沒有遇到任何問題。
6:32.460–6:33.360
一切都運作得很好。
6:33.440–6:34.680
那個小彈出視窗運作得很好。
6:34.860–6:40.720
我很喜歡你能看到自己說話時的文字,而不是
6:37.790–6:40.720
只看到那個小圖示在動。
6:40.820–6:43.320
因為有一半的時候,我會想:等等,它真的聽到我了嗎?
6:43.540–6:46.280
它連接到我在另一個房間的藍牙耳機了嗎?
6:46.540–6:48.100
比如說,這東西有在運作嗎?
6:48.460–6:51.200
在螢幕上看到文字也非常棒。
6:51.440–6:52.100
而且它是免費的。
6:52.320–6:53.360
我們有提到它是免費的嗎?
6:53.480–6:54.420
我很喜歡它是免費的。
6:54.540–6:55.640
我很喜歡它是開源的。
6:55.640–6:59.200
Fluid Voice 是我們 GitHub 儲存庫展示中眾多應用程式之一。
6:59.500–7:01.240
點擊這裡的連結,我們會展示其他應用程式。
7:01.620–7:02.020
在那邊見。
0:00.000–0:05.480
This little dictation app is free, open source, and
這個小小的聽寫應用程式是免費且開源的,並且
0:02.740–0:05.480
uses AI on your computer.
在你的電腦上使用 AI。
0:05.680–0:08.740
And I'm going to show you how you can install it and use it right now.
我將向你展示如何立即安裝並使用它。
0:09.360–0:12.360
Presented by Zapier, the AI automation company.
由 AI 自動化公司 Zapier 呈現。
0:12.920–0:15.660
Adam, everyone's using audio dictation apps on their computers.
亞當,現在大家都在電腦上使用音訊聽寫應用程式。
0:15.780–0:17.440
They're paying 15 bucks a month for it.
他們為此每月支付 15 美元。
0:17.500–0:20.180
I'm using Whisperflow, and we've got a better solution,
我使用的是 Whisperflow,而我們有更好的解決方案,
0:20.320–0:22.380
one that's completely free and works way better.
一個完全免費且效能更好的方案。
0:22.660–0:23.740
What do you use? What do you pay?
你使用什麼?你支付多少?
0:24.260–0:26.300
I use this thing called Super Whisper.
我使用一個叫做 Super Whisper 的東西。
0:26.460–0:28.980
It's using all the same models that all these other tools use.
它使用了與所有其他工具相同的模型。
0:28.980–0:34.580
And I'm paying, it's about $10 a month, but
我每月支付大約 10 美元,但
0:31.780–0:34.580
I don't like to think about it too much.
我不喜歡想太多。
0:34.960–0:36.160
10 is fine.
10 美元沒問題。
0:36.260–0:39.840
I pay 15 bucks a month because I pay monthly for this.
我每月支付 15 美元,因為我是按月付費。
0:40.000–0:41.760
I don't like it for a few reasons.
我不喜歡它有幾個原因。
0:41.900–0:43.780
Number one, I'm sending it up into the cloud,
第一,我將它上傳到雲端,
0:43.900–0:46.040
which I don't have any worries about any private information,
雖然我對私人資訊沒有擔憂,
0:46.120–0:48.720
but who knows who's going to subpoena what that I said, number one.
但誰知道誰會為了我說了什麼而發出法庭傳喚,這是第一點。
0:48.940–0:51.960
Number two, if it's going up into the cloud, it's taking longer.
第二,如果它要上傳到雲端,速度會比較慢。
0:52.280–0:55.360
And number three,
第三,
0:53.820–0:55.360
I happen to believe that there's a better user experience,
我認為有更好的使用者體驗,
0:55.440–0:57.240
and Whisperflow is not giving it to us.
而 Whisperflow 沒有提供給我們。
0:57.240–0:58.960
And neither are these other apps.
其他應用程式也沒有。
0:59.060–1:03.080
So you and I did the GitHub show, and
所以你和我們做了 GitHub 節目,並且
1:01.070–1:03.080
we talked about the top repos of the week.
我們討論了每週的前幾名程式庫。
1:03.160–1:07.880
And then I said,
然後我說,
1:04.733–1:06.307
I've got to atta
我必須
1:06.307–1:07.880
ch something to it that people are not starring enough,
為它添加一些人們不夠關注的功能,
1:08.340–1:09.400
and it's this.
就是這個。
1:09.560–1:11.060
It is something called Fluid Voice.
它叫做 Fluid Voice。
1:11.160–1:14.800
Look, it shot up after we did our GitHub show,
看,它在我们做了 GitHub 節目後迅速上升,
1:12.980–1:14.800
which is incredibly popular.
這非常受歡迎。
1:14.960–1:16.800
If you're not watching it, folks, you've got to watch it.
如果你還沒看,各位,你一定要看。
1:16.800–1:18.900
We reveal these types of things all the time.
我們經常揭露這類事物。
1:19.120–1:20.700
I'm one of those stars, Andrew.
我是那些星標之一,安德魯。
1:20.860–1:23.480
I'm that star right at the kind of top third there.
我就是那個位於頂部三分之一的星標。
1:23.600–1:26.040
I was like, oh, man, after I tried it and I started using it,
我想說,喔,天啊,在我嘗試並開始使用它之後,
1:26.240–1:29.640
I immediately came back to the repository, gave it a star,
我立刻回到該程式庫,給了它一個星標,
1:29.740–1:32.060
because I think these guys are going to be really successful.
因為我認為這些人會非常成功。
1:32.240–1:33.020
It's an awesome product.
這是一款超讚的產品。
1:33.140–1:37.200
It's fantastic because you get to customize it.
它很棒,因為你可以自訂它。
1:37.200–1:43.060
You get to download this or get the whole code base,
你可以下載這個或取得整個程式碼庫,
1:43.160–1:44.740
and you could change it to however you want.
然後按照你的喜好進行修改。
1:44.840–1:46.980
Or you could just go to their website, and you can get it.
或者你只需前往他們的網站,就可以取得它。
1:47.080–1:48.780
I got it and downloaded it for Mac.
我下載並安裝了 Mac 版本。
1:49.040–1:51.100
The setup is super-duper easy.
設定過程超級簡單。
1:51.580–1:54.820
Then once I had it, I got to pick which voice engine I wanted.
接著,當我安裝好之後,我必須選擇想要的語音引擎。
1:54.880–1:56.200
And I said, I want the blazing fast one.
我說,我想要那個速度飛快的那個。
1:56.280–2:00.040
You can go for whatever model you like.
你可以選擇任何你喜歡的模型。
2:00.040–2:05.560
And then you can customize it here by telling it what you want to use as your hot key,
然後你可以在這裡進行自訂,告訴它你想使用什麼作為熱鍵,
2:05.740–2:08.740
what you want to use as your continuation key.
你想使用什麼作為繼續鍵。
2:08.820–2:11.300
It will even do some edits on your text if you want it.
如果你希望,它甚至會對你的文字進行一些編輯。
2:11.540–2:15.220
Let me show you what it looks like compared to Whisperflow.
讓我展示一下它與 Whisperflow 相比的樣子。
2:16.000–2:18.120
On Whisperflow, I'm going to hit the FN key, and I'm going to say,
在 Whisperflow 上,我會按下 FN 鍵,然後我會說,
2:18.620–2:21.860
my name is Andrew, and
我的名字是安德魯,而且
2:20.240–2:21.860
I'm just going to be talking directly into the screen,
我會直接對著螢幕說話,
2:21.980–2:24.180
and I'm going to talk super fast because I'm a New Yorker at heart,
我會說得非常快,因為我骨子里是個紐約人,
2:24.260–2:25.400
and I will always talk super fast.
而且我總是說得非常快。
2:25.460–2:28.440
And you can see on the very bottom of my screen how all my WAV files are coming in.
你可以在螢幕最下方看到我所有的 WAV 檔案是如何進來的。
2:28.440–2:29.880
Let's see how long it takes for it to come through.
讓我們看看它需要多長時間才能處理完成。
2:30.800–2:31.180
Let go.
放開手。
2:31.440–2:31.920
It's thinking.
它正在思考。
2:32.420–2:33.360
Not bad, right?
還不錯,對吧?
2:34.240–2:34.420
Yeah.
是的。
2:34.680–2:34.840
Okay.
好的。
2:35.100–2:36.460
I think it basically got it.
我想它基本上已經處理好了。
2:36.680–2:40.340
Now I'm going to switch over to Fluid and watch this.
現在我要切換到 Fluid,請看。
2:41.720–2:44.840
My name is Andrew, and
我的名字是安德魯,而且
2:43.280–2:44.840
now I'm going to be using this Fluid voice thing.
現在我要使用這個 Fluid 語音功能。
2:44.920–2:49.580
And notice on the bottom of my screen,
注意看我的螢幕底部,
2:47.250–2:49.580
you can not just see that it's typing with the WAV thing,
你不僅可以看到它正在透過 WAV 進行輸入,
2:49.660–2:54.020
but it's also showing the words,
它還會顯示文字,
2:51.840–2:54.020
which lets me see that what I'm saying is actually going through.
這讓我看到我說的內容確實正在被處理。
2:54.240–2:58.840
And it's letting me see what kind of features I've got on and
它讓我看到我啟用了哪些功能,
2:56.540–2:58.840
where it's planning to paste it.
以及它計劃將文字貼到哪裡。
2:58.840–3:01.460
In this case, into Chrome.
在這種情況下,是貼到 Chrome 瀏覽器。
3:01.840–3:05.260
Let's let go of the button, the hotkey, and
讓我們放開按鈕,也就是熱鍵,然後
3:03.550–3:05.260
let's see how fast it pastes it in.
讓我們看看它貼上文字的速度有多快。
3:07.760–3:08.380
That's amazing.
這太驚人了。
3:08.440–3:08.660
Right.
沒錯。
3:09.120–3:11.240
And the fact that I can see it in real time is fantastic.
而且我能即時看到結果,這太棒了。
3:11.900–3:13.060
Really good piece of software.
這套軟體真的很好用。
3:13.060–3:15.920
I'm so glad that we got to share this with people.
我很高興我們能與大家分享這個工具。
3:16.420–3:17.720
What do you have to add on this?
你還有什麼要補充的嗎?
3:17.760–3:19.280
You've been using this a lot longer than I have.
你使用這套工具的經驗比我久多了。
3:19.360–3:20.700
I mean, there's so much.
我的意思是,這裡的功能太多了。
3:20.960–3:24.220
So number one, I think you glossed over something I think is really important here.
首先,我認為你略過了一個我認為非常重要的部分。
3:24.400–3:29.760
You know, Whisperflow, Super Whisper, all these other ones,
你知道的,Whisperflow、Super Whisper,還有其他這些軟體,
3:27.080–3:29.760
they're all companies that are charging us to use this software.
它們都是向我們收費以使用這些軟體的公司。
3:29.760–3:32.700
The models are either open or available.
這些模型不是開源的,就是可以取得的。
3:33.400–3:36.440
And these companies have packaged it into a nice Mac app for you.
這些公司將它們包裝成精美的 Mac 應用程式提供給你。
3:37.060–3:38.580
FluidVoice is open source.
FluidVoice 是開源的。
3:38.940–3:40.000
You could download the source.
你可以下載原始碼。
3:40.100–3:40.980
You could save it forever.
你可以永遠保存它。
3:41.140–3:42.920
You don't ever have to pay these guys money.
你永遠不需要付錢給這些公司。
3:43.260–3:46.160
Other people are going to come in here and contribute to it.
其他人會進來貢獻力量。
3:46.200–3:47.100
They're going to change it.
他們會修改它。
3:47.280–3:48.420
There's going to be versions of this.
會有這個軟體的不同版本。
3:48.420–3:52.900
The fact that this is open means that this is free and
它是開源的這個事實意味著它是免費的,而且
3:50.660–3:52.900
it's going to be free.
它將永遠免費。
3:53.240–3:54.760
And you get the control.
而且你擁有控制權。
3:54.900–3:57.820
Now, you don't have to customize it if you don't want to, but you could.
現在,如果你不想自訂,你不需要自訂,但你可以。
3:58.340–4:12.840
But what you've already pointed out is when
但你已經指出的重點是,當
4:01.240–4:04.140
you go into the settings, there's a lot of
你進入設定時,那裡有很多
4:04.140–4:07.040
customization available there, including, y
可用的自訂選項,包括,
4:07.040–4:09.940
ou know, running all of your transcriptions
你知道,在文字被貼上到應用程式之前,
4:09.940–4:12.840
through AI prompts before they get pasted into the application.
先透過 AI 提示進行所有轉錄。
4:13.020–4:17.580
And so you could have set up, I want a prompt for writing emails.
所以你可以設定,我想要一個撰寫電子郵件的提示。
4:17.580–4:20.080
Well, now it's going to remove your ticks.
現在,它會移除你的口頭語。
4:20.200–4:23.100
It's going to add, you know, a new line after Dear Andrew.
它會加入,你知道,在「親愛的 Andrew」之後換行。
4:23.500–4:25.080
I love the show that we do together.
我喜歡我們一起做的節目。
4:25.400–4:27.100
It'll add the new line there.
它會在這裡加入換行。
4:27.160–4:27.920
It'll add the comma.
它會加入逗號。
4:28.540–4:30.440
And you can customize those prompts.
你可以自訂這些提示。
4:30.800–4:32.300
You can also customize the dictionary.
你也可以自訂詞典。
4:32.680–4:42.240
I've already, when I get one of these tools,
我已經,當我拿到這些工具時,
4:35.070–4:37.460
the first thing I do is I drop into the cust
我做的第一件事就是進入自訂
4:37.460–4:39.850
om words and I add in all the names of the p
詞彙,並加入所有我
4:39.850–4:42.240
eople that I work with that I say out loud all the time.
經常大聲說出的同事的名字。
4:42.300–4:45.600
I add the names of the companies that I work with because
我會加入我合作的公司名稱,因為
4:43.950–4:45.600
they're all spelled weird.
它們的拼法都很奇怪。
4:45.600–4:53.980
And immediately, once you put them in this dictionary,
而且一旦你把它們放進這個字典,
4:49.790–4:53.980
again, fluid voice is going to use AI to plug in those words.
Fluid Voice 就會再次使用 AI 來填入這些單字。
4:54.380–5:00.700
And so it's never going to spell your name wrong when
所以當我對它說話時,它永遠不會拼錯你的名字,因為
4:57.540–5:00.700
I speak to it because it knows exactly how Andrew Warner is spelled.
它非常清楚 Andrew Warner 的拼法。
5:01.000–5:02.020
It knows how my name is spelled.
它知道我的名字怎麼拼。
5:02.120–5:04.400
My partner, Jesse Poogee, it knows all of that.
我的合夥人 Jesse Poogee,它都知道。
5:04.480–5:05.780
Is it J-E-S-S-I?
是 J-E-S-S-I 嗎?
5:05.900–5:06.860
J-E-S-I-E?
還是 J-E-S-I-E?
5:07.200–5:07.980
J-E-S-S-E?
或是 J-E-S-S-E?
5:08.160–5:09.680
Like fluid voice knows.
Fluid Voice 都知道。
5:09.680–5:11.940
So I've put all of that in place.
所以我把這些都設定好了。
5:12.200–5:13.080
And I agree.
我也同意。
5:13.160–5:14.740
I'm on the Blazing Fast model, too.
我也在使用 Blazing Fast 模型。
5:14.900–5:17.700
But I did actually test out a couple of the other models.
但我確實測試過其他幾個模型。
5:17.880–5:20.460
I tested one of the even faster ones.
我測試了一個更快的模型。
5:20.820–5:23.300
I also tested one that was, I don't remember which it was.
我也測試了一個……我記不起來是哪一個。
5:23.340–5:25.200
Yeah, the Cohere High Accuracy.
對,是 Cohere High Accuracy。
5:25.380–5:27.500
It got like two more accuracy points.
準確度只多了兩分。
5:28.360–5:29.740
But it's a little slower.
但速度比較慢。
5:30.200–5:31.860
And I decided, no, I can't tell a difference.
我決定,不,我感覺不出差別。
5:32.020–5:33.880
And so I'm on the same one that you're on.
所以我跟你用同一個。
5:34.420–5:39.140
The only weird thing, you know, when I first downloaded this,
唯一奇怪的是,你知道,我第一次下載這個的時候,
5:36.780–5:39.140
I installed it.
我安裝了它。
5:39.300–5:40.600
I clicked Blazing Fast.
我點擊了 Blazing Fast。
5:40.700–5:41.700
It started to install.
它開始安裝。
5:42.340–5:43.680
And then nothing happened.
然後就沒反應了。
5:43.860–5:44.680
And nothing happened.
一直沒反應。
5:44.880–5:46.540
And I actually wrote you a Slack message.
我甚至給你發了 Slack 訊息。
5:46.960–5:49.560
It was like, oh, man, Andrew, like, I was wrong.
我說:「喔,天啊,Andrew,我想我錯了。」
5:49.740–5:51.880
In the video, I said this is going to be amazing.
在影片裡我說這會很棒。
5:52.540–5:53.460
And it's not.
但事實並非如此。
5:53.540–5:54.060
I'm sorry.
對不起。
5:54.220–5:54.880
But you know what?
但你知道嗎?
5:54.900–5:56.200
I waited a couple more minutes.
我又等了幾分鐘。
5:56.200–5:57.520
The thing installed.
安裝完成了。
5:57.660–5:58.820
And it totally works.
它完全正常運作。
5:58.940–6:02.060
And I've already removed Super Whisper from my computer.
我已經從電腦上移除了 Super Whisper。
6:02.060–6:03.180
That's a good note.
這是個好提醒。
6:03.260–6:04.440
Because the same thing happened to me.
因為我也遇到了同樣的情況。
6:04.500–6:06.580
And because it happened to you, I knew to watch out for it.
因為你也遇到了,所以我會特別留意這個問題。
6:06.600–6:10.580
What happens is it will say 3% installed.
情況是它會顯示已安裝 3%。
6:10.760–6:11.980
And then it just stops.
然後它就停住了。
6:12.300–6:13.240
And it won't do anything.
什麼事都不會發生。
6:13.760–6:19.800
And then a few minutes later, maybe 10, 15 minutes later,
然後幾分鐘後,也許是 10 或 15 分鐘後,
6:16.780–6:19.800
it shifts from 3% to 98 to 100.
它會從 3% 跳到 98% 再到 100%。
6:20.020–6:20.860
And then it's done.
然後就完成了。
6:21.480–6:23.020
So that part wasn't working well.
所以那部分運作得不太好。
6:23.020–6:28.220
But little inconveniences to me like that are going to be worth the $10 a month.
但對我來說,像這樣的小不便,換取每月 10 美元的代價是值得的。
6:28.220–6:32.000
Plus, like, since it's installed and since it's running,
另外,既然它已經安裝並正在運行,
6:30.110–6:32.000
I've had zero issues.
我完全沒有遇到任何問題。
6:32.460–6:33.360
Everything has worked great.
一切都運作得很好。
6:33.440–6:34.680
The little pop-up works great.
那個小彈出視窗運作得很好。
6:34.860–6:40.720
I love that you can see the text as you're speaking instead
我很喜歡你能看到自己說話時的文字,而不是
6:37.790–6:40.720
of just the little thing that's moving.
只看到那個小圖示在動。
6:40.820–6:43.320
Because half the time, like, wait, can it actually hear me?
因為有一半的時候,我會想:等等,它真的聽到我了嗎?
6:43.540–6:46.280
Is it connected to my Bluetooth headphones that are in the other room?
它連接到我在另一個房間的藍牙耳機了嗎?
6:46.540–6:48.100
Like, is this working?
比如說,這東西有在運作嗎?
6:48.460–6:51.200
Seeing the text on the screen is amazing, too.
在螢幕上看到文字也非常棒。
6:51.440–6:52.100
And it's free.
而且它是免費的。
6:52.320–6:53.360
Did we mention that it's free?
我們有提到它是免費的嗎?
6:53.480–6:54.420
I love that it's free.
我很喜歡它是免費的。
6:54.540–6:55.640
I love that it's open source.
我很喜歡它是開源的。
6:55.640–6:59.200
Fluid Voice is one of many apps that we showed in the GitHub repo show.
Fluid Voice 是我們 GitHub 儲存庫展示中眾多應用程式之一。
6:59.500–7:01.240
Click the link here and we'll show you the others.
點擊這裡的連結,我們會展示其他應用程式。
7:01.620–7:02.020
See you there.
在那邊見。

影片筆記:Free, local audio dictation. Better than Wispr Flow, Superwhisper & Apple’s dictation

一句話總結

影片比較了付費雲端語音轉文字工具(Whisperflow、Super Whisper)與免費開源本地工具 Fluid Voice,強調 Fluid Voice 在隱私、速度、即時反饋及客製化(熱鍵、提示詞、詞典)上的優勢,並實測證明其優於現有付費方案。

核心重點

  • 隱私與成本:付費工具(如 Whisperflow)需將數據上傳雲端,存在隱私疑慮(可能被調閱 subpoena),且每月費用約 $10-$15 美元。
  • Fluid Voice 優勢
  • 免費且開源:無需付費,原始碼可下載修改。
  • 本地運行:數據不離開電腦,無雲端延遲。
  • 即時反饋:按下熱鍵後即時顯示文字,確認語音已接收並指定輸入目標(如 Chrome)。
  • 高度客製化:支援熱鍵設定、AI 提示詞(用於自動格式化或編輯)、自訂詞典(人名、公司名)。
  • 實測對比
  • Whisperflow:按下 FN 鍵後需等待雲端處理,底部僅顯示 WAV 檔案進程,速度較慢。
  • Fluid Voice:按下熱鍵後速度極快(Blazing fast),無需等待,即時顯示文字。
  • 安裝體驗:安裝進度條初期可能卡在 3% 數分鐘,隨後跳至 98-100% 完成,屬正常現象。
  • 模型選擇:推薦使用「Blazing Fast」模型;測試過「Cohere High Accuracy」發現準確度差異小但速度較慢,故未採用。

詳細大綱

1. 背景與贊助

  • 影片由 Zapier(AI 自動化公司)贊助。
  • 主題為電腦語音轉文字(dictation)應用程式的比較。

2. 現有付費工具的缺點

  • 隱私問題:數據上傳雲端,講者擔心資料被調閱(subpoena)。
  • 速度問題:雲端處理導致延遲。
  • 費用問題:每月約 $10-$15 美元。
  • 體驗問題:缺乏即時文字顯示與良好的使用者介面。
  • 提及工具:Whisperflow(Andrew 使用)、Super Whisper(Adam 使用)。

3. Fluid Voice 的功能與優勢

  • 核心價值:免費、開源(Open Source)、本地運行(On your computer)。
  • 技術基礎:使用與付費工具相同的模型(Models)。
  • 客製化能力
  • 可下載原始碼並修改。
  • 可設定熱鍵(Hot key)與繼續鍵(Continuation key)。
  • 可設定 AI 提示詞(AI prompts)進行文字編輯(例如自動格式化郵件)。
  • 可建立自訂詞典(Custom dictionary)以正確拼寫人名與公司名。
  • 使用者體驗:即時顯示正在輸入的文字,確認語音已接收並指定輸入目標(如 Chrome)。

4. 實測比較 (Demo)

  • Whisperflow 表現
  • 按下 FN 鍵說話。
  • 需等待雲端處理。
  • 底部僅顯示 WAV 檔案進程。
  • 速度較慢。
  • Fluid Voice 表現
  • 按下熱鍵說話。
  • 即時顯示文字與輸入目標。
  • 速度極快(Blazing fast),無需等待。

5. 安裝與模型選擇

  • 模型選擇
  • 推薦「Blazing Fast」模型。
  • 測試過其他模型(如 Cohere High Accuracy),發現準確度差異小但速度較慢,故未採用。
  • 安裝問題
  • 安裝進度條可能卡在 3% 數分鐘。
  • 隨後跳至 98-100% 完成。
  • 結論:儘管安裝過程有小不便,但免費與開源特性使其值得取代付費軟體。
  • Andrew 已移除 Super Whisper 改用 Fluid Voice。

6. GitHub Repo Show

  • Fluid Voice 是 GitHub 熱門倉庫之一。
  • 鼓勵觀眾觀看 GitHub repo show 以發現更多開源工具。

工具 / 模型 / 名詞整理

  • Zapier:AI 自動化公司,贊助本影片。
  • Whisperflow:付費語音轉文字應用程式,Andrew 使用的工具。
  • Super Whisper:另一款付費語音轉文字應用程式,Adam 使用的工具。
  • Fluid Voice(或 Fluid):免費、開源、本地運行的語音轉文字應用程式,被推薦的工具。
  • GitHub:程式碼託管平台,提及「GitHub show」與「top repos of the week」。
  • Mac:作業系統平台,Fluid Voice 可下載於此。
  • Chrome:瀏覽器,Fluid Voice 可將文字輸入至此。
  • Blazing Fast:Fluid Voice 中的模型選項名稱。
  • Cohere High Accuracy:Fluid Voice 中測試過的另一個模型選項。
  • WAV files:語音檔案格式,Whisperflow 底部顯示此類檔案進程。
  • FN key:Adam 提到按下此鍵來啟動 Whisperflow。
  • Hot key / Continuation key:Fluid Voice 的熱鍵與繼續鍵設定。
  • AI prompts:Fluid Voice 中用於文字編輯的提示詞功能。
  • Custom dictionary:Fluid Voice 中的自訂詞典功能。

操作流程整理

  1. 安裝 Fluid Voice
  • 從相關來源下載安裝。
  • 注意安裝進度條可能初期卡在 3%,隨後跳轉至完成,屬正常現象。
  1. 設定與客製化
  • 選擇模型:推薦使用「Blazing Fast」模型以獲得最佳速度。
  • 設定熱鍵(Hot key)與繼續鍵(Continuation key)。
  • (可選)設定 AI 提示詞(AI prompts)以進行自動格式化或編輯。
  • (可選)建立自訂詞典(Custom dictionary)以處理特定人名或公司名。
  1. 使用 Fluid Voice
  • 按下設定的熱鍵。
  • 開始說話。
  • 即時查看文字顯示與輸入目標確認。
  • 文字直接輸入至指定應用程式(如 Chrome)。
  1. 對比測試(若使用 Whisperflow)
  • 按下 FN 鍵。
  • 說話並等待雲端處理。
  • 觀察底部 WAV 檔案進程,體驗較慢的速度。

值得注意的限制或風險

  • 安裝進度異常:安裝過程中進度條可能卡在 3% 數分鐘,隨後才跳轉,可能讓使用者誤以為安裝失敗或卡住。
  • 模型選擇影響:若選擇非推薦模型(如 Cohere High Accuracy),可能導致速度變慢而準確度提升有限。
  • 平台限制:目前提及 Fluid Voice 可下載於 Mac 平台,需確認其他平台支援情況(影片主要聚焦於 Mac 體驗)。
  • 雲端隱私風險(針對競爭對手):使用 Whisperflow 等付費工具時,數據上傳雲端,存在被調閱(subpoena)的潛在隱私風險。

逐字稿辨識疑點

  • Fluid Voice / Fluid Voice thing:逐字稿中交替使用 "Fluid Voice" 與 "Fluid",以及 "Fluid voice thing"。
  • Jesse Poogee:Andrew 提到其夥伴的名字為 "Jesse Poogee",此拼寫較不常見,需查證是否為正確人名或口誤。
  • FN key:Adam 提到按下 "FN key" 來啟動 Whisperflow,通常 FN 為功能鍵,需確認是否為特定軟體的熱鍵設定或口誤。
  • Super-duper easy:形容詞疊用,可能為口語強調。
  • ticks:Adam 提到 AI 提示詞會 "remove your ticks",在文字編輯語境中,"ticks" 可能指 "ticks"(勾號)或為 "ticks"(口語中的停頓/雜音)或聽寫錯誤(如 "ticks" 應為 "ticks" 或特定標點符號),需查證。
  • 3% installed... stops... shifts from 3% to 98 to 100:安裝進度條的行為描述,雖為現象描述,但 "shifts" 一詞用於進度條跳變稍顯不尋常,可能為 "jumps" 或 "updates" 的口誤,但依要求保留原樣。
  • Andrew Warner:Andrew 自稱名字為 "Andrew Warner",需確認是否為講者真名或筆名。
  • J-E-S-S-I? J-E-S-S-I-E? J-E-S-S-E?:Andrew 拼讀夥伴名字時的疑問,顯示對名字拼寫的不確定性。

可延伸追問

  • Fluid Voice 在 Windows 或 Linux 平台上的支援情況如何?
  • AI 提示詞(AI prompts)的具體語法或範例為何?
  • 自訂詞典(Custom dictionary)的匯入/匯出格式為何?
  • Fluid Voice 的開源原始碼主要託管於哪個 GitHub 倉庫?
  • 「Blazing Fast」模型與其他模型在準確度上的具體數據差異為何?

尚未產生學習筆記

請在 Telegram 指令最後加上「學習」,例如:videonote 網址 英文 雙語 學習