全新Gemini功能详解——如何使用谷歌最新AI升级

translation of the given Markdown article into Chinese is as follows:

YouTube Transcript

语言: 英语（自动生成）（en）

[0:00] Google just delayed Gemini 3.5 Pro to

[0:02] June. Sundar walked on stage at

[0:04] presentation and literal])** "Give

[0:07] us until next month." And instead of

[0:09] Pro, they put Flash on the main stage.

[0:11] I've been running new Flash model on my

[0:12] Pro account for a week now across every

[0:14] surface Google gave us. The Geminiapp,

[0:17] AI studio,Workspace extensions, and

[0:20] native video drop. And here's my honest

[0:23] verdict. Flash quietly killed the Pro

[0:25] tier models. So, open the new Geminiapp

[0:27] interface and let's test it together.

[0:29] First up is multimodal vision in the

[0:31] Geminiapp. I uploaded a photo of my

[0:34] fridge, half empty, random ingredients,

[0:37] the kind of situation where you open it

[0:39] and star fish for 3 minutes. I asked

[0:42] it, "Look at everything in this fridge

[0:44] and give me a step-by-step dinner recipe

[0:47] using what's here. Then list anything

[0:49] I would need to grab from the store to

[0:50] complete it." watch what happens with

[0:52] the overlapping items in the back of the

[0:54] fridge. There are two jars partially

[0:56] hidden behind each other. Flash

[0:58] identifies both of them correctly and

[1:00] factors them into the recipe. That's not

[1:02] a trivial vision task. Most models

[1:04] either miss the hidden item or invent

[1:07] one that isn't there. The output gives

[1:09] you a full recipe with steps and a clean

[1:11] shopping list. It only includes the

[1:13] ingredients you're actually missing.

[1:15] Flash added nothing that wasn't actually

[1:17] there. This is a straight, usable

[1:19] answer. Multimodal in the Geminiapp has

[1:22] been solid for a while. Flash makes it

[1:24] faster and the output is easier than

[1:26] 2.5 Pro was on this same test. Quick

[1:30] pause while this thinking block keeps

[1:32] cranking because the pricing story is

[1:34] honestly the whole reason this video

[1:35] exists. If you're already paying for

[1:38] Gemini Pro, Claude, andChatGPT

[1:40] separately, you feel that power on your

[1:42] card every single month. So, here's why

[1:45] I want to tell you about our Pro plan on

[1:47] AI Master. The plan is built around one

[1:50] idea. You'll learn and you practice

[1:52] inside the same workspac

translated]) translationtranslated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation translated]) translation]) translation]) translation]) translation]) translation]) translation]) translation]) translation]) translation]) translation]) translation]) translation]) translation]) translation]) translation]) translation]) translation]) translation]) translation]) translation]) translation]) translation]) translation]) translation]) translation]) translation]) translation]) translation]) translation]) translation]) translation]) translation]) translation]) translation]) translation]) translation]) translation]) translation]) translation]) translation])** translation

Here's a quality translation of the given Markdown into Chinese:

[9:20] Here's a quick one before we get into

[9:22] 结构化数据。我记录了一个 chaotic

[9:25] 2分钟 voice memo on my phone.它 was

[9:27] first thing on a Monday morning. It was

[9:30] a real brain dump with no structure.

[9:32] Half thoughts about a project, a few

[9:34] customer follow-ups, and one idea I

[9:37] almost forgot, I dropped that audio file

[9:39] straight into the Geminiapp and asked

[9:42] it for a clean task list with

[9:43] urgentities. The response came back in

[9:46] about 10 seconds. Flash transcribed the

[9:48] whole memo first, then it grouped the

[9:51] related thoughts and ranked them by

[9:52] urgentities. There was no prep work at all

[9:55] on my end. If your morning usually

[9:57] starts in a notebook or a notesapp, try

[9:59] this workflow instead. talk for 2

[10:01] minutes during your morning drive, then

[10:03] let Flash structure it before you sit

[10:06] down at your desk. Here's another one

[10:08] that seems small, but solves a real

[10:10] daily problem.我 took a screenshot of an

[10:13] error message on my phone. It was one of

[10:15] those vague alerts that tells you

[10:17] nothing useful. I uploaded the

[10:19] screenshot into Gemini and asked it to

[10:21] explain what actually happened and what

[10:24] to do next. Flash read the screenshot

[10:26] and identified the root cause in plain

[10:28] language. Then it walked me through the

[10:30] exact fix step-by-step. There was no

[10:33] jargon and no digging through forum

[10:35] threads. The same workflow works forapp

[10:38] errors and for code errors in your

[10:40] terminal. If you've been Googling

[10:42] cryptic error messages for years, this

[10:44] one habit will save you real hours every

[10:47] single month. The next demo is JSON

[10:49] structures output in AI studio. I have

[10:52] 15 receipt photos from different

[10:54] countries in different languages and

[10:56] formats. The goal is to extract clean

[10:58] structures data from all of them without

[11:00] writing a single line of API code. I

[11:03] used to pay for a separate OCR API to

[11:05] handle exactly this kind of extraction.

[11:08] Flash does it in under 2 minutes

[11:10] natively in theui at no extra cost.

[11:12] That's the part that actually surprising

[11:14] him. In AI studio, there's a dedicated

[11:17] structures output panel where you define

[11:19] your schema visually. I set up fields

[11:21] for merchant name, date, total amount,

[11:23] currency, and line items. I did it all

[11:25] directly in theui without writing a

[11:27] single line of code. Then I uploaded all

[11:30] 15 receipts and ran the extraction in

[11:33] one request. The output is valid JSON

[11:36] for every receipt. Flash handles the

[11:38] language switching automatically. Every

[11:40] receipt, regardless of language, comes

[11:42] back in the same schema with the correct

[11:44] fields populated, and you can download

[11:46] the JSON file directly from AI studio

[11:49] with one click. If you do expense

[11:51] tracking or invoice processing, this is

[11:54] the workflow to use. This demo is

[11:56] work agometric chains in the Gemini

[11:58]app. You need work extensions

[12:00] enabled for this. Go to settings and

[12:01] then extensions. Turn on Google Drive,

[12:03]_docs, gmail, and calendar. Here's the

[12:06] single prompt I used. At Google Drive,

[12:08] find my May sales report. At Google

[12:10].Doc, create a new summary document from

[12:12] it. At gmail, drafting a team update with a

[12:15] link to that new document. watch Flash work

[12:18] through this. It opens drive and finds

[12:20] the file. Then it reads the content.

[12:22] Then it creates a new.Doc file and

[12:25] writes the summary. Then it drafts the

[12:27] gmail message with a link already in the

[12:29] body. It chain three separate Google

[12:31] products together from one prompt. I

[12:33] didn't touch a single one of those apps.

[12:35] This is the agometric Gemini moment that

[12:38] the I/O demos were building toward.

[12:40] You're not using Gemini as a text

[12:42] generator anymore. You're using it as an

[12:44] operator that moves through your actual

[12:46] tools and completes multi-step work on

[12:49] your behalf. Theapp mention syntax is

[12:52] how you control which tools it touches.

[12:54] Keep your prompt specific and it stays

[12:56] on track. okay, this is the section most

[12:59] reviews skip. Here are three things I

[13:01] found that Google is not putting in the

[13:04] heads. The first issue is long context

[13:06] context取. It got worse. Flash scores 7.6 points lower than Gemini 3.1 Pro on the MRCR V2])]) at 128,000 token context. MRCR measures how accurately a model retrieves specific information from a long document. Flash regression on it. The second issue is])])asticity. outputs got longer. Independent])-run shows Flash using twice the tokens on reasoning重 tasks compared to earlier models. I saw this consistently across my own tests. The third issue is the silent thinking level default change. When users migrated from 2.5 Pro to 3.5 Flash, the default thinking level in the Geminiapp dropped from high to medium. Google did not announce this change. A lot of users are encountering]))**.

[13:56]legate])** Leone star $

[13:58] Leone star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star star

[14:43] see you in the next one.