Name: GPT Image 1.5 vs Nano Banana Pro 誰更快？Gemini 3 Flash 速度實測！ (We test GPT Image 1.5 vs Nano Banana Pro with Gemini 3 Flash speed)
Uploaded: 2025-12-18T14:09:46.000Z
Duration: 6 min 52 s

形容詞比較級

Maybe this is a little more interesting because this happens.

或許這樣會比較有趣，因為會發生這種情況。

These kind of upgrades happen less, at least as far as OpenAI is concerned.

這種升級比較少見，至少就 OpenAI 而言是這樣。

過去簡單式

And so an upgrade to OpenAI's capabilities around generating images.

這代表 OpenAI 在生成圖像方面的能力有所提升。

I think clearly this is, you know, really aimed at.

我覺得這很明顯，你知道，是針對。

Nanobananapro is an elephant in the room that is really sucking all the oxygen out because it's damn good, by the way.

Nanobananapro 這個東西真是個大問題，因為它太好用了，把所有人的注意力都吸走了。

And so they want to position this is better in all the same ways.

所以他們想把這個定位成在所有方面都更好。

You know, faster, better control, improved text rendering.

你知道，更快、更好的控制、改進的文字渲染。

Now they're all kind of finally getting good at this.

現在大家好像都終於在這方面做得不錯了。

They say four times faster than the previous model.

I played around with it and it may be faster, but

我試玩了一下，它可能比較快，但是

I wouldn't say it's Gemini fast as far as that's concerned.

就這方面來說，我不會說它有 Gemini 那麼快。

I wanted to like, just because we've got a show and because I like show and tell, kind of put a single prompt through both of these and see how they do.

因為我們有節目，而且我喜歡展示和分享，所以我想試著對這兩者下一個提示，看看它們的表現如何。

So I've got Gemini open, I've got Nano Banana Pro.

所以我打開了 Gemini，還有 Nano Banana Pro。

Turn this article text into an infographic image because I love playing around with the infographics.

把這篇文章的文字變成一張資訊圖表，因為我超愛玩資訊圖表的。

This is just a random post I found on theaibreak substack.com There you got a magical plug tutorial 10 notebook lm use cases that will TedX your productivity.

這是我從 theaibreak substack.com 上隨便找到的一篇文章。這裡有篇「魔法外掛教學：10 個能讓你的生產力 TEDx 化的小筆記本 LM 用例」。

Okay, so I'm going to fire this off in Gemini.

好，那我就用 Gemini 來跑跑看。

I'm doing the fast, which we're going to talk about in a second.

我用的是快速版，等一下會再聊到。

And then I'm going to fire it off in ChatGPT and kind of see what it comes back with.

然後我會用 ChatGPT 來跑跑看，看看它會回什麼。

Generally, Gemini, I mean, that was really fast.

總體來說，Gemini，我的意思是，它真的很快。

Gemini came back in like 10 seconds, maybe, maybe that.

Gemini 大概 10 秒就回覆了，也許，也許就這麼快。

然後你得到了穩定的文字呈現。

我們來看看在這裡有沒有道理。

一、二、三、三、五、四、六、六、十。

Okay, so not all of them are numeric brain Braum.

好的，所以不是全部都是數字大腦。

Okay, 12, but be that as May 10 use cases, that is 12 of them.

好的，十二個，但就這樣吧，十個使用情境，也就是十二個。

I'm going to go ahead and do a new one and I'm not going to do fast because I think when I tested this initially, I wasn't testing with fast.

我準備要來做一個新的，而且我不會用快的，因為我覺得我一開始測試的時候，我沒有用快的來測試。

我記得我當時是用 pro 來測試的。

Okay, so ChatGPT did not do what I wanted it to do.

好的，所以 ChatGPT 沒有做到我想要它做的事。

It gave me a description here Is a ready to use.

它在這裡給我一個描述，說「準備好了」。

Sometimes it needs that explicit instruction, I've noticed.

我注意到，有時候它需要明確的訓示。

Otherwise it's like, yeah, here's what I'd do.

不然它就會說，「嗯，這是我會做的」。

If I create an image, it's like, yeah, just do that.

如果我建立一個圖片，它就會說，「嗯，就這樣做」。

This is probably the most exciting audio podcast to ever listen to.

這大概是有史以來最令人興奮的音頻播客了。

一、二、三、四、五、六、七。

Actually 10 use cases, writing and font consistency and everything looks.

其實有 10 個使用情境，寫作和字體的一致性等等看起來都很棒。

There's no made up AI alien words here, so that's good.

這裡沒有亂編的 AI 外星語，這點很好。

So from a speed perspective, definitely a little slower.

所以從速度來看，確實是比較慢一點。

我喜歡這個節目裡面的這種競賽的感覺。

Sometimes demos like this work and sometimes I regret it afterwards where I'm like, oh, that was only really interesting if you were actually watching it.

有時候像這樣的示範會成功，有時候事後我會後悔，覺得「喔，這只有在你實際看的時候才真的有趣」。

For the audio podcast listeners, while this is working, why don't we talk about the other thing and then we'll check back in on ChatGPT.

對於聽音頻podcast 的聽眾，在它運作的時候，我們不如聊聊別的事情，然後再回來看看 ChatGPT。

Because this is an indication of how slow the image model is for ChatGPT.

因為這顯示了 ChatGPT 的圖像模型有多慢。

But we do also have Google announcing this morning the release of Gemini 3 Flash, which is, as I kind of showed off very briefly earlier, it's faster model.

不過今天早上 Google 也宣佈了 Gemini 3 Flash 的發布，正如我剛才簡短展示過的，它是一個更快的模型。

Essentially they say it's a solid match or beats 2.5 pro on multimodal reasoning coding tool use runs up to three times faster.

基本上他們說它在多模態推理、編碼工具使用方面，能與 2.5 Pro 匹敵或超越，速度快上三倍。

And yeah, it's basically the default that you're going to find in AI mode in the Gemini app and all that kind of stuff.

而且，基本上它就是你在 Gemini app 和其他類似東西的 AI 模式裡會找到的預設選項。

Looks absolutely nothing like our site, but it did.

看起來跟我們的網站完全不一樣，但它確實做到了。

And it also, by the way, it did not get our images correct.

而且順帶一提，它也沒有正確辨識我們的圖片。

It got our names, but these are definitely not us.

它辨識出了我們的名字，但這些絕對不是我們。

But it, it turned this around in 10 seconds.

但它，它在 10 秒內就完成了這件事。

This, this was made in 10 seconds coded, which is pretty inside the Evolution blue.

這個，這個是在 10 秒內完成的程式碼，這真的太進化了。

Do I need to change, you know how we market this podcast?

我需要改變，你知道我們怎麼行銷這個 podcast 嗎？

A lot of the AI kind of, you know, imagery and stuff is dystopian and steampunk.

很多 AI 的影像和東西都帶有反烏託邦和蒸汽龐克的風格。

So anyways, for what it's worth, 10 seconds inside a Gemini 3 flash and it created an interactive site.

總之，不管怎麼說，在 Gemini 3 的 flash 裡頭 10 秒鐘就建立了一個互動式網站。

I wonder if like some of these other tabs actually take me to places or if it's just the landing page.

我很好奇，像這些其他的分頁，會不會真的帶我去別的地方，還是隻是登陸頁面而已。

Looks like it's probably just the landing page right now, but.

看起來現在大概只是登陸頁面，不過。

Okay, let's check back in on the infographic.

一、二、三、四、五、六、七、八、九。

字幕列表影片播放

GPT Image 1.5 vs Nano Banana Pro 誰更快？Gemini 3 Flash 速度實測！ (We test GPT Image 1.5 vs Nano Banana Pro with Gemini 3 Flash speed)

stuff

perspective

essentially

aspect