节目 学英语,听播客 下一集
第 28 集

SeedRealtime: When AI Watches, Listens, and Talks

SeedRealtime:当 AI 边看、边听、边说
approximately 7 minutes B1
音频已上线 点击收听
Mike

Welcome back to "Learn English with Podcasts"! Sarah, quick question - have you ever pointed your phone at something and wished it could tell you what it sees?欢迎回到"Learn English with Podcasts"!Sarah,问你个问题——你有没有过把手机对着某个东西,希望它能告诉你它看到了什么?

Sarah

Ooh, like when I look at a menu or a sign I can't understand? That would be so useful.哦,比如盯着看不懂的菜单或路牌的时候?那也太实用了。

Mike

Exactly. And there's good news. A company just released a new AI model called SeedRealtime. It can watch, hear, and talk all at the same time.没错。而且有好消息。一家公司刚发布了新模型,叫 SeedRealtime。它能同时看、听、说。

Sarah

Wait, I thought voice assistants could already do that. I talk, they answer.等等,我以为语音助手早就能做到了。我说话,它们回答。

Mike

That's the big difference. Old assistants work like a relay race - turn your voice into text, then think about it, then turn the reply back into sound. Each step adds delay, and each step loses a little meaning.这正是最大的不同。老助手像接力赛跑——把你的声音变成文字,再思考,再把回答变回声音。每一步都加一点延迟,每一步都丢一点信息。

Sarah

A relay race! So it's like a game of telephone, where mistakes get passed along?接力赛!所以就像传话游戏,错误会一路传下去?

Mike

Exactly! SeedRealtime does all of this inside one model - video, sound, and words together. Nothing gets lost, and nothing gets slow. The full-duplex part means both sides can talk at the same time, like a real call.完全正确!SeedRealtime 把所有环节放在一个模型里——画面、声音、文字一起处理。什么都不丢,也不会卡顿。"全双工"就是双方可以同时说话,像真的打电话一样。

Sarah

Okay, so what can I actually do with it? Give me a real example.好,那我到底能拿它做什么?给我举个真实例子吧。

Mike

Picture this. Two travelers sit down in a Chinese restaurant. The menu is all in Chinese, and they have no idea. They point the camera at the menu and ask, "What should we order?"想象一下。两个游客走进一家中餐馆。菜单全是中文,他们完全看不懂。他们拿起手机对着菜单,问"我们该点什么?"

Sarah

And the AI actually looks at the dishes and explains in English? Like it understands that "yuxiang pork" has no real fish inside?然后 AI 真的看着菜,用英语解释?比如它知道"鱼香肉丝"里并没有真正的鱼?

Mike

Right! And here's the kicker - it can explain why. Because "yuxiang" means the flavor is fish-style, not that fish is in it. It connects what it sees with what it knows.对!而点睛一笔是——它还能解释"为什么"。因为"鱼香"指的是鱼的香味,不是说食材里有鱼。它把看到的画面和它懂的知识连起来。

Sarah

So it's more than a translator. It's like a knowledgeable friend at the table. What else?所以它不只是翻译,更像一个懂行的朋友同桌而坐。还有呢?

Mike

Try a museum. You say, "please tell me when you see the golden tiger." The model keeps watching the video as you walk - and when the statue appears, it speaks up by itself.再比如博物馆。你说"看到金虎雕像的时候提醒我"。模型在你走动时一直盯着画面,一旦雕像出现,它会自己开口。

Sarah

Wait, so it doesn't wait for me to ask each time? It watches on its own and acts on its own?等等,所以它不用我一次次提问?它自己盯着看,自己行动?

Mike

That's the proactive part. It tracks what's happening and decides, "this is worth saying now." And nobody asked it to.这就是"主动"的部分。它一直观察变化、自己判断"现在值得说"。而且不是被任何人提醒的。

Sarah

But what about talking over each other? Would it jump in while you're still speaking?那抢着说话怎么办?你正在说的时候,它会突然插进来吗?

Mike

It has a great sense of rhythm. It knows when to wait, when to pause, and when to chime in. And it's trained to avoid talking over people - it only speaks at the right moment.它节奏感很好。它知道什么时候该等、什么时候该停、什么时候该插嘴。它被训练成不要抢话,而是自然地接在合适的空当里。

Sarah

Okay. That beats every group chat I've ever been in. What about noisy places?好吧,这比我待过的所有群聊都强。那在嘈杂的环境里呢?

Mike

It's tested in crowded cafés, busy streets, even with a TV on in the background. It learned to focus on your voice. So if people nearby are chatting, it won't start replying by accident.它专门在拥挤的咖啡馆、繁华的街道、甚至开着电视的背景里测试过。它学会了只锁定你的声音。所以就算旁边的人闲聊,也不会被误触发、乱接话。

Sarah

So it's polite, patient, and focused. When can I try it?所以它又礼貌、又有耐心、又专注。我什么时候能试试?

Mike

It's live right now in the Doubao app. You open a call, and the AI talks back while watching what you show it. No typing, no waiting.它现在已经在豆包 App 上线了。你发起通话,AI 一边看你展示的东西,一边回你话。不用打字,不用等待。

Sarah

This is the future. So what's the one thing you want listeners to take away?这就是未来啊。所以你想让听众记住的最重要的一句话是什么?

Mike

Here it comes - the old AI answered your words. The new one sees your world. It doesn't just understand what you say; it sees where you are. That changes everything.关键来了——过去的 AI 只会回应你说的话。新的 AI 能看到你生活的世界。它不只是理解你说的话,而是看得见你在哪里。这改变了一切。

Sarah

Mike, every episode I learn something, and this one's my new favorite. See you next time!Mike,每期我都能学到新东西,这期是我最喜欢的。我们下期见!

Mike

Ha, my pleasure! Thanks for watching "Learn English with Podcasts" - and next time, bring your camera.哈哈,这是我的荣幸啊。感谢收听"Learn English with Podcasts"——下次来,记得带上你的摄像头。

已复制