Betting on the Wave

Hi everyone, good afternoon. Welcome to our podcast.

大家好，下午好，欢迎来到我们的播客。

And today we're really happy to welcome my co-host, Chang, from Hockey Stick Growth.

今天我们非常高兴请到我的联合主持人陈畅（Chang），他来自 Hockey Stick Growth。

And we're also really happy to welcome the first guest of our podcast — Alex, the CEO and founder of Higgsfield.

我们也很高兴迎来本播客的第一位嘉宾——Alex，Higgsfield 的 CEO 兼创始人。

And maybe Chang will also give an introduction, and then introduce Alex to our audience.

稍后陈畅也会做个介绍，把 Alex 正式介绍给各位听众。

Yeah, I know that you've had a very successful career and journey so far.

是的，我知道你到目前为止有一段非常成功的职业生涯和经历。

And your last company, AI Factory, was a big hit, and it got successfully acquired by Snapchat.

你的上一家公司 AI Factory 非常成功，后来被 Snapchat 顺利收购。

And you became an executive at one of the most successful tech companies.

之后你成了全球最成功的科技公司之一的高管。

And now that you're starting your next journey, it's already become a super hit — everyone's talking about how you reached 50 million in ARR in just six months.

而如今你开启了新的征程，一上来就大获成功——大家都在谈论你如何在短短 6 个月里做到 5000 万美元的 ARR。

We're super excited to have you today, and to really deep dive into your successful journey, as well as what's next for Higgsfield in the future.

今天非常激动能请到你，想好好深入聊聊你这段成功的历程，以及 Higgsfield 接下来要做什么。

Thank you so much.

非常感谢。

I've spent my whole professional career — like 10 years — working in the space of video AI.

我整个职业生涯——差不多 10 年——都在做视频 AI 这件事。

And it's exciting to see how we progressed from images 64 pixels by 64 pixels to, let's say, 15-, 20-, 30-second-long consistent videos.

看着我们从 64×64 像素的图像，一路走到如今 15 秒、20 秒、30 秒的连贯视频，真的很令人兴奋。

Gotcha, gotcha. So for many early-career professionals, they think: if I can work for Snapchat, it'll be my dream job.

明白明白。很多处在职业早期的人会想：要是能进 Snapchat 工作，那就是我的梦想工作了。

What did you see in the market that became the biggest trigger for you to leave a well-paid and well-respected job in the tech industry to start your own journey?

那你在市场上看到了什么，成了你离开这份高薪又受人尊敬的科技行业工作、出来自己创业的最大触发点？

What we've seen is that social media became the main distribution platform — the most important form of media.

我们看到的是，社交媒体已经成了最主要的分发平台，是最重要的媒介形态。

We've seen how mobile phones and advanced cameras really propelled — and then TikTok — they propelled a certain type of content.

我们看到智能手机和高级摄像头是怎么把某一类内容推起来的——后来 TikTok 又进一步把这类内容推向高峰。

But that represents only maybe 5 or 10 percent, because the content on social media today is not diverse.

但这其实只占到大概 5% 到 10%，因为今天社交媒体上的内容并不多样。

The diversity of content — if we look at TV, or if we look at Hollywood — is way higher compared to social media.

内容的多样性——你看电视，或者看好莱坞——都比社交媒体高得多。

And I think, naturally, video AI is closing the gap.

而我认为，视频 AI 正在很自然地把这个差距补上。

In the past, any work that would require 3D animation or VFX would be impossible to do for social media, because of the high cost of production.

过去，任何需要 3D 动画或视觉特效（VFX）的工作，对社交媒体来说都不可能完成，因为制作成本太高。

And social media really requires velocity of releasing — a few pieces of content a day.

而社交媒体真正需要的是发布速度——一天得发好几条内容。

And I think that's where gen AI provides that new benefit.

我觉得这正是生成式 AI 带来的新价值所在。

Yeah, and maybe — for the audience who aren't that familiar with our product, Higgsfield —

对，也许——给那些现在还不太熟悉我们产品 Higgsfield 的听众——

could you give us a short intro about what Higgsfield actually is?

你能不能简单介绍一下 Higgsfield 到底是什么？

And what are some core products that you think will quickly give our audience or users an aha moment?

还有，你觉得哪些核心产品能很快让我们的听众或用户产生"啊哈"时刻？

What kind of use cases are they creating most frequently right now?

他们现在最常用它来做哪些用例？

Totally. Where it all started for Higgsfield, when we launched on March 30th, was camera control effects.

完全可以。Higgsfield 一切的起点，是我们 3 月 30 日上线时推出的"相机控制特效"（camera control effects）。

It became extremely popular across various use cases, especially in the e-commerce space.

它在各种用例里都极受欢迎，尤其是在电商领域。

And that's where we see consistent usage over time — all the e-commerce use cases drive a lot of adoption.

这也是我们看到使用量长期保持稳定的地方——所有电商用例带来了大量采用。

And then I'd say we probably haven't really seen yet the adoption of longer-form AI-generated videos.

而另一方面我得说，长视频形态的 AI 生成内容，目前可能还没真正被大规模采用。

We haven't.

我们还没看到。

We have not, we have not.

还没有，还没有。

I think Sora gave us an early preview of what it will look like.

我觉得 Sora 给了我们一个早期预览，让我们看到它大概会是什么样子。

But the perception was that it's slop — brain rot, and so on.

但大家的印象是它很"slop"（粗制滥造）——属于"脑腐"（brain rot）那一类。

And so, for example, at Higgsfield, after we successfully raised our Series A, we can invest substantially more in professional workflows, and really focus on these high-value use cases.

所以拿 Higgsfield 来说，在顺利完成 A 轮融资之后，我们可以在专业工作流上投入多得多的资源，真正聚焦这些高价值用例。

Mm-hmm. Right.

嗯，对。

Then are you implying that right now a lot of AI video adoption is still in a more amateur user group — more casual use cases?

那你的意思是不是说，现在很多 AI 视频的使用还集中在比较业余的用户群体里——更多是随性的用例？

Not necessarily the professional video producers, for longer-form video and more commercial use cases?

而不一定是那些专业视频制作者，去做更长的视频和更偏商业的用例？

Yeah. Most of the usage comes from professionals — way over 60 percent, based on our user survey.

对。大部分使用都来自专业用户——根据我们的用户调研，远超 60%。

And what's interesting to me is that all these people are interested in video AI, and Higgsfield is probably the easiest platform to access all those various models.

让我觉得有意思的是，这些人都对视频 AI 感兴趣，而 Higgsfield 大概是接触所有这些不同模型最便捷的平台。

As the number of people interested in video AI grows exponentially, we expect to see the same growth for Higgsfield, as we ride the wave.

随着对视频 AI 感兴趣的人数呈指数级增长，我们预期 Higgsfield 也会随之同步增长——我们就是在骑这股浪潮。

But that's only one part of the story.

但这只是故事的一部分。

The second part is trying to go another route, and really upgrading existing workflows.

第二部分，是换一条路走，真正去升级现有的工作流。

For example, storyboards and these use cases have been around for quite a while.

比如，故事板（storyboard）和这类用例其实已经存在很久了。

Clearly it can be done with AI.

显然，这些都可以用 AI 来完成。

But it requires more than just a model integration — it requires a collaboration layer and so on.

但它需要的不只是接入一个模型——它需要一个协作层（collaboration layer）等等。

And that's what we're building in Higgsfield as well.

而这也正是我们在 Higgsfield 里构建的东西。

Nice. So you have the new features you announced last week — about popcorn.

很棒。你们上周还发布了新功能——关于 popcorn。

Yes, and the team plan releases in the next 48 hours.

对，而且团队版（team plan）会在接下来 48 小时内发布。

Right, I have one more question, maybe to go deeper into the user profile.

对，我还有一个问题，想再深入聊聊用户画像。

Did you see any change in the users — in the content creation or generative media group — in how they're evolving?

你有没有看到用户——在内容创作或生成式媒体这个群体里——在如何演变？

Because at the beginning, I think a lot of users came from image generation, and then came Sora 1 or other video models.

因为一开始，我觉得很多用户是从图像生成来的，后来出现了 Sora 1 或其他视频模型。

And some users evolved from image creation to video creation, and also brought in some new users or new professional users.

然后一些用户从做图进化到做视频，同时也带来了一些新用户、新的专业用户。

So have you seen a similar trend in Higgsfield, and how does your product try to evolve to meet the different needs of different groups of new users?

那你在 Higgsfield 有没有看到类似的趋势？你们的产品又是怎么演进，去满足不同新用户群体的不同需求的？

That's a good question.

这是个好问题。

We don't focus on animated content that much, while Midjourney clearly does.

我们不太聚焦动画类内容，而 Midjourney 显然在做这块。

We're focused on the use cases for social media, which really require a lot of static photo shoots — let's say with AI actors — and also product placements.

我们聚焦的是社交媒体的用例，它们真正需要的是大量静态拍摄——比如用 AI 演员（AI actors）——以及产品植入。

I think that's most of the adoption on the platform, driven by these two major segments.

我觉得平台上大部分采用，都是由这两大板块驱动的。

OK. And on the product side, the next question is — there are different kinds of directions I see on the Higgsfield platform.

好的。在产品这一侧，下一个问题是——我在 Higgsfield 平台上看到了好几种不同的方向。

Because, for one, you have different features or tools based on image, and then different tools or features based on video.

因为，一方面你有基于图像的不同功能或工具，另一方面又有基于视频的不同工具或功能。

And then you create different use cases — like the app, popcorn, storyline.

然后你又据此做出不同的用例——比如 app、popcorn、storyline。

So I just wonder, could you share some of the philosophy behind this — your product roadmap?

所以我很好奇，你能不能分享一下这背后的一些理念，以及你的产品路线图？

And how did it go from zero to the current products offered on the platform?

它又是怎么从零，走到今天平台上这些产品的？

Look, the bet we're making is: we'll build the best product, the best interface, for social media professionals to access video AI tools.

你看，我们下的赌注是：我们要为社交媒体专业人士打造最好的产品、最好的界面，让他们用上视频 AI 工具。

You're right — as models improve, the product must drastically improve.

你说得对——随着模型变强，产品也必须大幅进化。

So I guarantee you, the product in one month will look completely different than it does today.

所以我向你保证，一个月后的产品会和今天完全不一样。

And we're ready for this race.

我们已经为这场竞赛做好了准备。

Because we expect that by the end of the decade, most of the content on socials is gonna be generated by AI.

因为我们预期，到这个十年结束时，社交媒体上大部分内容都会由 AI 生成。

Probably it's gonna happen by the end of 2028, in the next three years.

这很可能在 2028 年底、也就是未来三年内就会发生。

And when we think about that, we're really talking about adoption of maybe 50 to 100 million users.

而当我们这么想的时候，谈的其实是大概 5000 万到 1 亿用户的采用规模。

Clearly now, most companies like Higgsfield are at maybe the high hundreds of thousands of paying users.

而现在很明显，大多数像 Higgsfield 这样的公司，付费用户大概还在几十万这个量级的高位。

That's still two or three magnitudes lower than it will be.

这仍然比它将来的规模低了两到三个数量级。

So clearly, today we're going through the early-adoption stage.

所以很清楚，我们今天还处在早期采用阶段。

And we do believe that by the end of next year, there are gonna be one or two major use cases that'll be considered fully covered by AI.

我们确实相信，到明年底，会出现一两个被认为"已被 AI 完全覆盖"的重大用例。

Like, for example, in LLMs, everyone's talking about how customer support is being replaced by AI.

就像在大语言模型领域，大家都在说客服正在被 AI 取代。

And I think the same — we're gonna see at least one killer use case next year.

我觉得视频也一样——明年我们至少会看到一个杀手级用例。

No one knows what it's gonna be.

没人知道它会是什么。

The structure of the e-commerce market is also not balanced.

电商市场的结构也是不均衡的。

Meaning there are large marketplaces that can impose their own agenda, and they have a lot of control.

也就是说，有一些大型平台能够强加自己的规则，掌握很大的话语权。

And there are all these smaller direct-to-consumer brands — think about creative teams of 3 to 6 people who use Excel today.

同时还有大量更小的 DTC（直接面向消费者）品牌——想想那些今天还在用 Excel 的 3 到 6 人创意团队。

And when the shakeout happens, it's important to acknowledge there's a major part of the market that's programmatic.

而当洗牌发生时，有一点很重要：市场里有很大一部分是程序化（programmatic）的。

So many companies — TikTok, Meta, Google — they're all built on programmatic advertisements.

TikTok、Meta、Google 这么多公司，全都建立在程序化广告之上。

The algorithm.

算法。

Yes, so it's a really high-margin business.

对，所以这是个利润率非常高的生意。

So I think the most profound application of video in social media is actually going to touch all these ad-tech platforms.

所以我认为，视频在社交媒体上最深远的应用，最终会冲击到所有这些广告技术（ad-tech）平台。

Meaning they found some equilibrium, where, let's say, the user experience on Instagram is good enough — every eight or ten videos, we watch ads.

意思是，它们找到了某种平衡——比如说，Instagram 的用户体验"足够好"：每看八条或十条视频，我们就看一次广告。

And users think it's good.

用户觉得这样挺好。

But it also makes enough money for Meta to pay a billion dollars for each researcher, right?

但同时它也赚到了足够多的钱，让 Meta 能给每个研究员开出十亿美元，对吧？

And I think at some point, with gen AI, that equilibrium is gonna look different.

而我认为，到某个时点，随着生成式 AI 的到来，这个平衡会变得不一样。

So there are gonna be winners — likely Google and ByteDance — but there are gonna be some losers, and Meta can be the loser as well.

所以会有赢家——很可能 Google 和字节跳动是赢家——但也会有输家，而 Meta 也可能成为那个输家。

OK, yeah, yeah.

嗯，好，好。

That's a very bold prediction.

这是个非常大胆的预测。

Yeah, I would say so.

对，我会这么说。

It's actually one of the most popular views right now — and that's why it's being discussed in Silicon Valley.

其实这现在是最流行的观点之一——所以硅谷都在讨论它。

Yeah, there's definitely a possibility of this happening, right?

对，这件事确实有可能发生，对吧？

I think it's unfair to say today that Google or ByteDance are at risk — simply because their distribution is strong.

我觉得，今天就说 Google 或字节跳动处于风险之中并不公平——单是它们的分发能力就很强。

Their vertical integration capabilities are strong, the teams are strong, and they have the capacity.

它们的垂直整合能力很强，团队很强，也有足够的资源储备。

So I wouldn't bet against these two.

所以这两家，我不会赌它们输。

But everyone else — I mean Meta, Snapchat, Pinterest — that's a way bigger market, which is gonna be impacted by video as well.

但其他所有人——我是说 Meta、Snapchat、Pinterest——那是一个大得多的市场，同样会被视频冲击。

You have a user base now, and you receive product requests from them — to make new twists on your products.

你现在有了用户基础，会不断收到他们的产品需求——希望你给产品做出新花样。

And then you have your own core product roadmap, right?

同时你又有自己核心的产品路线图，对吧？

Because you have to ship features, like, every two weeks, based on your schedule.

因为按你的排期，你差不多每两周就要发一批新功能。

So how do you balance the feature requests from users against your own vision, or your core product roadmap?

那你整体上是怎么在"用户提的需求"和"你自己的愿景／核心路线图"之间做平衡的？

Especially in such a dynamic environment.

尤其是在这么动态多变的环境里。

Yeah, my main learning is to focus less on my own vision, and really focus on what users need today.

嗯，我最大的心得是：少盯着自己的愿景，真正去关注用户今天需要什么。

And just accept it's gonna change.

并且坦然接受它一定会变。

But the reality is that those professional users — if they pay for video gen AI today, they'll basically be paying for video gen AI forever.

但现实是，那些专业用户——如果他们今天为视频生成式 AI 付费，基本上就会一直付下去。

Because they see this as the future.

因为他们认定这就是未来。

So we just need to learn how to adapt to their needs, which are shifting as well.

所以我们只需要学会去适应他们的需求，而这些需求本身也在不断调整。

To be fully clear, I think nine out of the top ten agencies use Excel today.

把话说透一点：我认为头部十大代理商里，有九家今天都在用 Excel。

And today they use it for storyboarding.

而且他们今天就是用它来做故事板（storyboard）。

There are maybe only a few large ads where Excel was involved on a daily basis — but storyboarding happens constantly; like, every minute someone makes a storyboard for a Higgsfield or commercial purpose.

可能只有少数几支大广告会天天动用 Excel——但故事板是时时刻刻在发生的；几乎每分钟都有人在为某条 Higgsfield 或商业用途做故事板。

And then, naturally, the industry is going to evolve from storyboards to social media ads.

然后，行业会很自然地从故事板演进到社交媒体广告。

It's gonna start with social media ads, maybe in emerging markets where risks are lower, then it's gonna go to the United States.

它会从社交媒体广告开始，可能先在风险较低的新兴市场，然后再走向美国。

And then it's gonna get amplified by performance-marketing budget, and the whole space is going to evolve, and the use cases are going to evolve.

接着它会被效果营销（performance marketing）的预算放大，整个赛道会演进，用例也会演进。

So the goal for us is to make sure that the Higgsfield products evolve.

所以我们的目标，就是确保 Higgsfield 的产品跟着演进。

When I talk about the team plans, it's a relatively easy bet — it's not that much added complexity for existing users.

我说的团队版（team plan），其实是个相对容易下的赌注——它对现有用户没有增加太多复杂度。

Single-player mode and multiplayer mode are kind of two different branches.

单人模式和多人协作模式，算是两条不同的分支。

But when I look at our main page today, I can fully acknowledge that we're actively A/B testing different variations.

但当我今天看我们的主页时，我完全承认：我们正在积极地对不同版本做 A/B 测试。

Users got attracted to Higgsfield because of the visual effects and all these camera effects I was talking about.

用户最初被 Higgsfield 吸引，靠的是我前面说的那些视觉特效和相机特效。

I think the main page should evolve as well.

我觉得主页本身也应该跟着进化。

OK.

好的。

Yeah. And in terms of verifying how well you're actually satisfying users' requests — do you use any systems or frameworks to decide?

嗯。那在"验证你到底有没有满足好用户需求"这件事上——你们有没有用什么系统或框架来判断？

We're actively monitoring the ratings — the number of complaints on Zendesk.

我们在密切监控评分——以及 Zendesk 上的投诉数量。

We actively monitor refund rates via Stripe.

我们通过 Stripe 密切监控退款率。

The number of support emails — we track this very actively on a weekly basis, broken down by categories.

客服邮件的数量——我们每周都很认真地追踪，并按类别拆分。

And I can tell you that today, all these AI video tools, including Higgsfield — the main issue is the cost.

我可以告诉你，今天所有这些 AI 视频工具，包括 Higgsfield——最主要的问题是成本。

It's cost. It's the cost.

是成本，就是成本。

Because it ultimately comes down to this: the entry price, just to try things out in mature markets like the US, is maybe 30 dollars a month.

因为归根结底是这样：在美国这样的成熟市场，光是想试一试，入门价大概是每月 30 美元。

And still, with Sora, the cost of a good-looking video might be 3 to 5 dollars, after several attempts to get something good.

即便如此，用 Sora，一条好看的视频，在反复尝试几次、做出像样的东西之后，成本可能要 3 到 5 美元。

And then someone could try to make a more sophisticated video with Sora, and it could be hard to pull off.

而如果有人想用 Sora 做一条更复杂的视频，那可能就很难搞定。

That's where they just run out of credits.

结果他们的额度（credits）就用光了。

They get their initial aha moment — where they see their idea visualized, which was never possible before.

他们得到了最初那个"啊哈"时刻——看到自己的想法被可视化出来，这在以前是不可能的。

But then the people who want to push it further may feel that it just becomes too expensive and too difficult.

但接下来，那些想再往前推进的人，可能会觉得它实在太贵、也太难了。

I think most people have this initial aha moment with Higgsfield, but then they run out of credits, and they don't get the exact piece they're comfortable posting.

我觉得大多数人在 Higgsfield 上都有这个最初的"啊哈"时刻，但随后额度用光，又做不出那条让自己满意、敢发出去的成品。

So improving the quality, and investing more into R&D and post-training to get more predictable outputs — I think that's a top priority.

所以提升质量、加大在研发和后训练（post-training）上的投入、得到更可预测的输出——我认为这是头号优先级。

Continuity of the video, of the content — right.

视频、内容的连续性——对。

I probably have one more question on the product.

关于产品，我大概还有一个问题。

We discussed — you mentioned a lot about short drama in your previous podcast.

我们之前聊过——你在上一期播客里讲了很多关于短剧（short drama）的内容。

And you have a really full view on how this type of content evolved, especially in recent years.

而且你对这类内容是怎么演变的，尤其是近几年，有一套相当完整的看法。

So does that actually feed into your decisions or vision for the product roadmap — especially something like popcorn, where you have storyline?

那这些有没有真正影响到你在产品路线图上的决策或愿景——尤其像 popcorn，你们还有 storyline？

I think when it gets longer, you could probably create a short drama.

我觉得当它做得更长时，大概就能做出一部短剧了。

...around the short-drama market — I learned a lot, and I'm happy to share that.

……围绕短剧市场，我学到了很多，也很乐意分享。

Clearly, over the last century, the most popular type of content consumed is drama.

很明显，过去一个世纪里，被消费得最多的内容类型就是剧（drama）。

Dramas on TV, dramas, books.

电视剧、剧集、小说。

And it was a huge gap that, years ago, no one did dramas for the phone.

而曾经有一个巨大的空白：很多年前，没人为手机做剧。

And that trend started from China.

而这个趋势是从中国开始的。

Now this market is over 10 billion dollars globally.

如今这个市场在全球已经超过 100 亿美元。

And it grew — over Tencent — and it was less than a billion, I think, five years ago, so it grew very, very quickly.

它一路增长——经由腾讯（Tencent）等渠道——而我记得五年前它还不到 10 亿美元，所以增长非常非常快。

And when you go to airports here in the United States, most of the books are gonna be drama-related.

你看，在美国的机场，大部分书都会是和"剧情"相关的。

You go to Kazakhstan, same story.

你去哈萨克斯坦，也是一样的情况。

It's sort of the type of content that travels really well, and that can be monetized really well.

它是那种能很好地跨地域传播、也能很好地变现的内容类型。

When I say monetized really well, I really want to double down on the demographics.

我说"变现能力很强"，是想特别强调它的人群结构（demographics）。

Because dramas tend to relate to older demographics, and especially the fictional dramas.

因为剧往往和更年长的人群相关，尤其是虚构题材的剧。

So the fictional dramas — the ones about werewolves, vampires — have a lot of opportunity to grow, due to video AI capabilities.

所以这些虚构题材的剧——关于狼人、吸血鬼的那些——借助视频 AI 的能力，有很大的增长空间。

Because typically these dramas are made with budgets 10 to 100 times lower than a traditional TV budget.

因为这类剧的制作预算，通常比传统电视剧低 10 到 100 倍。

So we're talking about a few thousand dollars a minute.

我们说的是每分钟几千美元。

Clearly, no 3D animation can be done under that budget.

显然，在这个预算下根本做不了 3D 动画。

So that's where video gen AI provides essential capabilities.

所以这正是视频生成式 AI 能提供关键能力的地方。

Although this is not where the disruption is gonna come from.

不过，颠覆并不会从这里发生。

I think the first area for disruption is actually gonna be related to the promotion of these dramas.

我认为第一个被颠覆的环节，其实会和这些剧的"推广"有关。

These drama companies grow heavily through performance marketing.

这些短剧公司，主要靠效果营销（performance marketing）来增长。

Right. Yeah. Yeah.

对，是的，是的。

They also have organic distribution channels, where they cut trailers.

他们也有自有的（organic）分发渠道，会在那上面剪预告片。

Yes, yes.

对，对。

The most exciting part of the drama, the trailer — yes, the trailer — on the social media channels.

短剧里最抓人的部分就是预告片（trailer）——对，就是预告——放在社交媒体渠道上。

Some drama companies sell from their own organic accounts.

有些短剧公司会用自己的自有账号来卖。

So that can also potentially be one of the channels they use — but they spend a lot on performance marketing.

所以这也可能是他们用的渠道之一——但他们在效果营销上花了很多钱。

Right, right. Which helps make variations of the top-performing creative.

对，对。它能帮你把那些表现最好的创意素材做出各种变体。

And that's a day-and-night difference for performance marketing.

而这对效果营销来说，是天壤之别。

When you think about it, the budget to make a good drama can be 500K.

你想想看，做一部好短剧的预算可能是 50 万美元。

Maybe one out of seven, one out of ten dramas must be a hit.

大概每七部、每十部里得有一部是爆款。

But then the budget to promote it can be tens of millions of dollars.

但用来推广它的预算，可能高达数千万美元。

It's way, way bigger.

要大得多、大得多。

So there's gonna be just one creative that takes most of this budget.

所以最后会有那么一条创意素材，吃掉这笔预算的大头。

So that's why I say this is a major opportunity.

这就是为什么我说这是个重大机会。

And this is the first area. I think the second area is really going to come from outside of them — sort of a bottom-up approach.

这是第一个领域。我认为第二个领域，真正会从他们之外冒出来——算是一种自下而上（bottom-up）的路径。

We've seen some companies, like Fox News, partner with a short-form drama company recently and invest in them.

我们看到一些公司，比如 Fox News，最近和一家短剧公司合作并投资了它。

But that's really a bit of a bottom-up approach.

但那确实有点像自下而上的打法。

I think what we're gonna see is that the way funding in TV and Hollywood works is being changed as well.

我觉得我们会看到的是，电视和好莱坞的投资方式也正在改变。

As everyone tries to rely on well-performing IP, so they can guarantee some return on capital.

因为所有人都想押注表现稳定的 IP，好确保一定的资本回报。

Ultimately, there's a lot of IP that was popular in the 80s and 90s.

归根到底，有大量在 80、90 年代很火的 IP。

And today, one of the popular topics I constantly hear about is doing reskinning.

而今天，我反复听到的一个热门话题，就是做"换皮重制"（reskinning）。

Mm-hmm. Yeah, yeah, yeah, yeah.

嗯，对对对对。

So, it's not exactly video-to-video.

所以，它并不完全是"视频到视频"（video-to-video）。

But it's sort of video-to-video.

但它又有点像"视频到视频"。

So it's not just applying an anime style and calling it good.

它不是简单地套一个动漫风格，就说"挺好"。

It's not like what I was doing with face filters in the past.

它不像我过去做人脸滤镜（face filters）那样。

It's really about: okay, what if we take Mickey Mouse from the 1960s — which is even now in the public domain —

它真正在做的是：好，如果我们拿 1960 年代的米老鼠（Mickey Mouse）——它现在甚至已经进入公有领域（public domain）——

and what if we just iterate over different visual styles?

然后用不同的视觉风格去反复迭代，会怎么样？

And the same can be applied to taking old movies and trying to make the next chapter of the movie.

同样的思路也可以用在旧电影上，去尝试做出这部电影的"下一章"。

And sort of do this unification, or — how to say — make older actors young again.

再做一种统一处理，或者说——把上了年纪的演员重新变年轻。

So, yes. And change different details — so it's presented in a different world.

对对。再改一些不同的细节——让它呈现在一个不同的世界里。

Yeah, yeah, yeah.

对，对，对。

So the IP is well trusted, and they just make different skins.

因为 IP 本身就很受信任，他们只需要做出不同的"皮肤"（skins）。

That's popular in games like Counter-Strike — even these hyper-casual games, they all typically monetize reskinning really well.

这在游戏里很常见，比如 Counter-Strike——甚至那些超休闲（hyper-casual）游戏，通常也都能靠换皮很好地变现。

Same machine, same engine.

同一套机制，同一个引擎。

And I think we'll see some of that by the end of the decade as well.

我觉得到这个十年结束时，我们在内容上也会看到类似的现象。

There are gonna be new AI-first franchises, and those franchises will likely start from something that was already battle-tested before.

会出现新的 AI 原生（AI-first）系列 IP，而这些系列很可能会从某个早已被市场验证过的东西起步。

I see, I see. Great.

懂了，懂了，很好。

And we're probably gonna dive into growth and marketing.

接下来我们大概会深入聊聊增长和营销。

Because, as you mentioned, Higgsfield just had such a hit this year, and went to 50 million ARR really quickly.

因为正如你提到的，Higgsfield 今年特别火，很快就做到了 5000 万美元 ARR。

So, for part of our audience — Higgsfield actually entered a very noisy market at the beginning, because there were already a lot of video-generation or image-generation tools in the market.

那对一部分听众来说——Higgsfield 其实一开始进入的是一个非常嘈杂的市场，因为市面上已经有很多做视频生成或图像生成的工具了。

So could you share the story of the cold start?

所以你能不能分享一下"冷启动"的故事？

At the very beginning, how did you actually grow from one feature into the market — or what kind of experiments did you try to cold-start the platform?

在最最开始，你是怎么从一个功能切入、一步步做到市场里的——或者说，为了冷启动这个平台，你们尝试了哪些实验？

I think our approach doesn't really change.

我觉得我们的打法其实没怎么变。

We see there are two types of creators.

我们看到创作者分两类。

The first type are those who want to make just viral content.

第一类是那些只想做爆款内容的人。

And the second type are educators — educators are the most popular type of key opinion leaders.

第二类是教育者（educators）——教育者是最受欢迎的一类 KOL（关键意见领袖）。

For us — and I can tell you this is the same story we saw at Snapchat — the natural interest to adopt AI visual tools is just very high.

对我们来说——我可以告诉你，这和我们在 Snapchat 看到的是同一个故事——大家采用 AI 视觉工具的自发兴趣本来就非常高。

And these accounts grow.

而这些账号会不断成长。

So partnerships with those accounts, those influencers, those creators, actually help build awareness about what we're building at Higgsfield.

所以和这些账号、这些网红、这些创作者建立合作，确实能帮我们打响 Higgsfield 在做什么的认知度。

And also, at the beginning — to dive deeper — which channels do you find these two groups of people are most active on?

嗯。还有，一开始要做这件事——再深入一点——你发现这两类人最活跃在哪些渠道上？

Or which types of channels have you tried?

或者说，你们都试过哪些类型的渠道？

Instagram is the most productive.

Instagram 的产出最高。

Instagram is the most productive, and YouTube is next.

Instagram 产出最高，其次是 YouTube。

OK, OK.

好的，好的。

But you're also super successful on Twitter — so how do you position that?

但你们在 Twitter 上也非常非常成功——那你是怎么给它定位的？

Twitter is a great platform, but it has its own thing — it has very well-defined subcommunities.

Twitter 是个很棒的平台，但它有自己的特点——它有非常清晰界定的子社区（subcommunities）。

Each subcommunity could be a little extreme, in terms of how far they push.

每个子社区都可能有点极端，在它们各自"用力"的方向上。

But it also includes a lot of professionals, obviously.

当然，它里面也包含了大量专业人士。

And Instagram is just a little more — if you think of it as a growth machine, and really making sure there's a well-systematized process, and you appeal to a wider audience, not just each of these small subcommunities — Instagram is definitely that platform.

而 Instagram 则更适合——如果你把它当作一台增长机器，真正去建立一套体系化的流程，去触达更广的受众，而不只是这些小子社区——那 Instagram 绝对就是那个平台。

And as Higgsfield builds more video-workflow functionality, then YouTube becomes more important as well.

而随着 Higgsfield 构建出更多视频工作流的功能，YouTube 也会变得更重要。

And you're also talking about the e-commerce companies, the different marketing teams.

你刚也提到了电商公司、不同的营销团队。

So how do you think about distribution on LinkedIn, and more on the B2B side of partnerships?

那你怎么看 LinkedIn 上的分发，以及更偏 B2B 的合作这一侧？

Yeah, I think LinkedIn comes after YouTube.

嗯，我觉得 LinkedIn 排在 YouTube 之后。

So we're still not necessarily in a position to go after specific use cases yet.

所以我们其实还没到能去主攻具体垂直用例的阶段。

Right now we cover those workflows for creative marketing teams first — that's why presence on Instagram and YouTube is very important.

现在我们先覆盖创意营销团队的那些工作流——这就是为什么在 Instagram 和 YouTube 上的存在感非常重要。

And once we feel the technology has matured, and hopefully the platform is just as mature, then focusing on specific use cases is going to be important.

等我们觉得技术已经成熟了，平台本身但愿也同样成熟时，再去聚焦具体的用例就会变得重要。

Got you, got you.

懂了，懂了。

So from the very beginning, you're more aiming for a really broad-appealing type of brand campaign.

所以从一开始，你更倾向于做那种受众面很广的品牌营销。

So that the early adopters of AI technology — especially AI video technology — get familiar with your brand.

好让 AI 技术——尤其是 AI 视频技术——的早期采用者，先熟悉你们的品牌。

And because they're naturally early adopters, it'll be a lot easier for them to start adopting.

而因为他们天生就是早期采用者，让他们上手采用就会容易得多。

Before you start to have more verticalized operations targeting very specific use cases in marketing.

在这之后，你们才开始做更垂直化的运营，去瞄准营销里非常具体的用例。

Exactly. Because it becomes more and more verticalized, like, every month.

正是如此。因为它差不多每个月都在变得越来越垂直化。

But it's important to acknowledge that storyline is the only model today to make the final video.

但有一点很重要：今天 storyline 是唯一能做出"成片"的方式。

It can be done with YouTube, popcorn, and other models as well — but I think it's still a little bit early.

用 YouTube、popcorn 和其他模型也能做——但我觉得现在还有点早。

Clearly, social media professionals don't need to make those small segments; they need the final video.

很明显，社交媒体专业人士不需要做那些小片段；他们要的是最终成片。

But the technology is not fully there yet.

但技术还没完全到位。

It's coming in the next three months.

它会在接下来三个月里到来。

Again, the Higgsfield platform is gonna look completely different.

再说一遍，Higgsfield 平台到时会变得完全不一样。

I just think we kind of follow the natural evolution of the technology.

我觉得我们就是顺着技术的自然演进在走。

Cool, cool. And for all your peer competitors that are also following the same technology trend —

很好，很好。那对所有同样在追这股技术趋势的同行竞争对手来说——

what would you say is the biggest thing that sets you apart?

你觉得让你们最与众不同的那一点是什么？

How come you stand out, while there are other competitors that may or may not make it?

为什么你们能脱颖而出，而其他那些竞争对手却不一定能？

Yeah, like OpenAI — they build their own social media.

嗯，比如 OpenAI——他们在做自己的社交媒体。

I think Luma is a very talented team of researchers; they'll probably go after TV and Hollywood.

我觉得 Luma 是一支非常有才华的研究团队；他们大概会去攻电视和好莱坞。

In the social media space, yes, there are companies that try to take a different approach.

在社交媒体这块，确实有公司在尝试不同的路径。

We're just taking the approach of riding the wave, and betting that the wave is gonna get bigger.

我们的打法就是骑在浪潮上，并赌这股浪潮会越来越大。

And we can be on top of the wave, really riding it, right? That's our bet.

而我们能站在浪头上，真正去骑它，对吧？这就是我们的赌注。

And we stay within the realm of social media marketing — it's just gonna get more verticalized by use case as we move forward.

我们守在社交媒体营销这个范畴里——随着往前走，它只会按用例越来越垂直化。

The storyboard is just the first example of that; the camera control was example number zero.

故事板只是其中的第一个例子；相机控制是第 0 号例子。

The storyboard is number one.

故事板是第 1 号。

And you're gonna see a way higher pace of iteration by use case from Higgsfield going forward — and this is a function of both the technology maturing and the market maturing.

往后你会看到 Higgsfield 按用例迭代的速度快得多——而这是技术成熟和市场成熟共同作用的结果。

And to deep-dive on social media — you just mentioned influencer marketing.

再深入聊聊社交媒体——你刚提到了网红营销（influencer marketing）。

I've heard a lot of stories that influencers or KOLs are getting really competitive right now, because everyone wants to chase the top KOLs or top creators.

我最近听到很多说法，说现在网红或 KOL 的争夺非常激烈，因为大家都想去抢头部 KOL、头部创作者。

That's a very good point.

这是个非常好的点。

What's actually surprising to me is the number of people who want to make mind-blowing content.

真正让我意外的，是想做出惊艳内容的人数。

It's crazy. They want to make it, the Higgsfield platform is good for that, they just do it, and it goes viral.

多到夸张。他们想做，Higgsfield 平台正好适合，他们一做，就火了。

So that number is definitely not small.

所以这个数字绝对不小。

But I think the question is about the educators.

但我觉得真正的问题在于教育者（educators）。

And I can tell you it's like less than fifty.

而我可以告诉你，这个数字大概不到 50 个。

Oh, okay.

哦，好。

Yeah, that's definitely a problem.

对，这绝对是个问题。

Because the adoption of AI — especially in the marketing space — has pros and cons.

因为 AI 的采用——尤其在营销领域——有利也有弊。

The pro is that there are a lot of experimental budgets.

好处是，这里有大量可用于尝试的预算。

The con is that people kind of love to be taught — they love the workshops and so on.

坏处是，人们其实挺喜欢被人教的——他们喜欢工作坊之类的形式。

But the number of people who can actually do that is very small.

但真正能去教这件事的人，非常少。

And right now you're seeing a lot of companies — brands big and small — all trying to pitch the top AI educators.

而现在你能看到很多公司——大大小小的品牌——都在争着去对接头部 AI 教育者。

So what's your suggestion to your peer founders who are also trying to pitch the same group of creators?

那你对那些同样想去对接同一批创作者的同行创始人，有什么建议？

I think that's definitely a limitation.

我觉得这肯定是个局限。

I see some companies trying to develop their own AI spokesperson — growing their own accounts becomes more important.

我看到一些公司在尝试打造自己的 AI 代言人——养自己的账号变得更重要了。

But that's definitely, if we look at this...

不过这当然……如果我们这么看的话……

It takes longer to develop your own account.

养自己的账号要花更长时间。

Yeah, exactly.

对，正是如此。

And I think qualified educators is one of the major concerns holding the industry back.

我觉得"够格的教育者"是拖住整个行业的主要瓶颈之一。

I think the other trend I see is that, alongside the top KOLs, there are also educators who want to grow themselves into top KOLs.

我看到的另一个趋势是：除了头部 KOL，还有一些教育者想把自己做成头部 KOL。

So do you have a strategy, or a way, to develop this partnership — or ecosystem — to help those educators, or those people in the middle layer, get involved and grow into advocates talking about you?

那你有没有什么策略或办法，去发展这种合作——或者说生态——帮那些教育者、那些处在中间层的人参与进来，并成长为愿意谈论你们的人？

It's just difficult for us to engage the maybe 2,000 creators we work closely with — it's hard to fully personalize the process.

对我们来说，要服务好那大概 2000 名跟我们紧密合作的创作者，是很难的——很难把整个流程完全个性化。

We're actively working on that, but I'm telling you, it's not easy to personalize a community of like 2,000 creators all over the world.

我们在积极地做这件事，但说实话，要把一个分布在世界各地、约 2000 人的创作者社区个性化运营，并不容易。

We're actively experimenting with that.

我们正在积极地做这方面的实验。

I see. Another question — could you share some metrics on how you actually evaluate the performance of those influencers?

明白。再问一个——你能不能分享一些指标，看你们到底是怎么评估这些网红的表现的？

Especially today, in such a noisy market, what kind of metrics do you think are still very important to evaluate their performance, or to continue the partnerships?

尤其在今天这么嘈杂的市场里，你觉得哪些指标在评估他们的表现、或决定是否继续合作时，依然很重要？

Yeah, that's a good point.

嗯，这是个好点。

I think you always want to measure the performance of posts over time — your post versus competitors' posts, and so on.

我觉得你总是想衡量帖子随时间的表现——你的帖子对比竞品的帖子，等等。

And that's all well-defined: you look at CPM, cost per engagement, and try to measure which channel is the most productive and efficient.

而这些其实都有成熟定义：你看 CPM、每次互动成本（cost per engagement），试着衡量哪个渠道产出最高、效率最高。

But I'd say the most challenging thing is dealing with bots.

但我得说，最棘手的事情是应付机器人（bots）。

Bots are a real issue.

Bot 是个实实在在的问题。

A lot of engagement numbers are fake. Just fake.

很多互动数据都是假的。就是假的。

Right. Impression numbers too.

对。曝光数据也是。

Impressions for sure, but engagement as well.

曝光肯定是，但互动也一样。

In the past, it was easy to look at the engagement-to-views ratio and say whether it's fake or not.

过去，你很容易通过"互动/播放比"判断它是不是刷的。

Now both engagement and view numbers can be fake.

现在，互动数和播放数都可以是假的。

And that, by the way, is the major problem for the whole industry.

顺带说一句，这是整个行业的大问题。

I think that's even a way bigger problem than Higgsfield.

我觉得这甚至比 Higgsfield 自身的问题大得多。

You go to Snap, you go to Meta, you go to any other platform — you see some number of views, but you can never verify if that's real.

你去 Snap，去 Meta，去任何一个平台——你会看到一个播放数字，但你永远没法验证它是不是真的。

So I think that's the overall problem of trust in the numbers presented on social media.

所以我认为，这是社交媒体上"呈现出来的数字"整体的信任问题。

And we hope that gen AI is gonna become the new form of media.

而我们希望，生成式 AI 会成为新的媒介形态。

It's hard to say this today, but over time things are gonna get personalized — and I hope this will eventually bring more transparency.

今天说这话还很难讲清楚，但随着时间推移，一切会变得越来越个性化——我希望这最终能带来更高的透明度。

And there will be more horizontal solutions across all these social media platforms, to drive a higher level of transparency.

而且会出现更多横向（跨平台）的解决方案，贯穿所有这些社交媒体平台，去推动更高的透明度。

Today it's a major issue — for everyone: even for bigger brands, for smaller companies, for up-and-coming companies like Higgsfield, it's the same issue.

今天这是个大问题——对所有人都是：无论是大品牌、小公司，还是像 Higgsfield 这样正在崛起的公司，面对的都是同一个问题。

Yeah, and which platform would you say is doing better at fighting the bots, providing more parity?

嗯，那你觉得在打击机器人、提供更真实数据方面，哪个平台做得更好？

I know that when I was at Snap, it was a top priority.

我知道在我还在 Snap 的时候，这是头等大事。

I know a lot of people like to say that Twitter isn't doing a great job, but I don't have any recent research, so I'd reserve my opinion.

我知道很多人喜欢说 Twitter 在这方面做得不好，但我没有最新的研究数据，所以我先保留意见。

OK. And you just mentioned the creators community.

好。你刚提到了创作者社区。

I think the creators community is a really important part of how you foster the ecosystem.

我觉得创作者社区是你培育生态非常重要的一环。

Yet I think the users community is also very important — I see you have this 'core' for the community.

不过我觉得用户社区同样很重要——我看到你们为社区做了一个"核心"（core）。

So I just wonder if you have any best practices for building the user community?

所以我想问，你有没有什么构建用户社区的最佳实践？

Because everyone knows you can put UGC content on the website for everyone to see — that's what we see — but how do you actually build up a community?

因为大家都知道，你可以把 UGC 内容放到网站上供所有人看——这是我们能看到的——但你到底是怎么把一个社区做起来的？

Yeah, at Higgsfield, we have several tiers of communities and creators.

嗯，在 Higgsfield，我们把社区和创作者分成了好几个层级。

It comes with pros and cons.

这有利也有弊。

The con is that when some creators learn they're not in the top-tier community, they get unhappy.

弊端是，当一些创作者发现自己不在最高层级的社区里时，他们会有点不高兴。

But I think we also try to support a lot of external communities, competitions, and so on.

但我觉得我们也在努力支持大量外部社区、各种比赛等等。

Hmm. OK, hmm. OK.

嗯，好，嗯，好。

I think — to that point — it's like trying to upgrade a mature creator to a middle-level creator.

我觉得——就这一点而言——它有点像是把一个成熟的创作者，往上提升成一个中间层级的创作者。

And then incentivize them to make a good-looking AI-generated video, which may actually inspire others to try video AI.

然后激励他们做出一条好看的 AI 生成视频，这条视频反过来可能会激发别人去尝试视频 AI。

And I think being an educator is the best job today.

而我认为，做一名教育者，是今天最好的工作。

Probably, like, the AI video tutors or mentors on the internet right now —

大概就像现在网上那些 AI 视频的导师或讲师——

they're making different kinds of videos using different tools, and also tutorials or workshops, for people who want to quickly learn the AI tools.

他们用各种不同的工具做各种视频，还做教程或工作坊，给那些想快速学会 AI 工具的人。

Yeah, the first time I got to learn about Higgsfield was actually through a creator.

对，我第一次知道 Higgsfield，其实就是通过一位创作者。

The creator had a portfolio on Higgsfield.

这位创作者在 Higgsfield 上有一个作品集（portfolio）。

And when we were looking for UGC creators, that creator sent over their portfolio, saying, "Hey, these are all the AI videos I generated."

当时我们在找 UGC 创作者，他就把作品集发过来，说"嘿，这些都是我用 AI 生成的视频"。

So in terms of the user community, do you see it more as a way to engage creators and existing users?

所以就用户社区而言，你更多是把它看作一种让创作者和现有用户保持活跃的方式？

Or do you also see it as a different way to grow and expand?

还是说，你也把它看作一种不一样的增长、扩张方式？

Yeah, I think we're just constantly working on it.

嗯，我觉得我们一直在不断打磨这件事。

I think that's the most difficult thing — how to personalize and build a fair process for a few thousand people.

我觉得最难的就是——怎么为几千个人做到个性化、又建立一个公平的流程。

That's just difficult.

这确实很难。

As a creator, especially when it's public, unfair treatment or evaluation can really spark a lot of controversy.

作为创作者，尤其当一切都是公开的时候，不公平的对待或评价真的会引发很大的争议。

I think it's still a little bit of a niche community.

我觉得它现在还算是个有点小众的社区。

Gotcha, I see.

懂了，明白。

And among all the go-to-market channels — we talked a lot about creator marketing, and touched a little on the user community —

那在所有的市场进入（go-to-market）渠道里——我们聊了很多创作者营销，也稍微提了用户社区——

are there any other channels you prioritize at the moment?

你目前还会优先考虑哪些其他渠道吗？

Just getting as much educational material out as possible.

就是尽可能多地把教育性的素材发出去。

Got you. Got you.

懂了，懂了。

And also — maybe this is my personal question too — have you started to try anything in terms of GEO?

还有——这也许也算我个人的问题——你们有没有开始在 GEO 方面做一些尝试？

Like, what we call making your products or content appear more in ChatGPT or those chatbots — have you tried something like that?

就是我们说的，让你的产品或内容更多地出现在 ChatGPT 或那些聊天机器人里——你们有没有试过类似的事？

Yeah, I think, not yet.

嗯，我觉得还没有。

I mean, I know this is going to grow.

我是说，我知道这块一定会增长。

Our priority is just to cover all of it — making sure that whatever social media marketing workflow exists today, and is growing, Higgsfield is part of that.

我们的优先级就是把它整个覆盖住——确保今天存在、并且在增长的任何一种社交媒体营销工作流里，都有 Higgsfield 的身影。

On the AI video generation process — we are not building our own ad platform.

在 AI 视频生成这条链路上——我们不会去做自己的广告平台。

I think it's important for us to become the backbone for social media video generation using generative AI.

我觉得对我们来说重要的是：成为用生成式 AI 做社交媒体视频生成的"骨干"（backbone）。

Hmm. OK. Hmm.

嗯，好，嗯。

This is a bold idea.

这是个大胆的想法。

Basically you're talking about how the different social media platforms will change.

你基本上是在讲不同的社交媒体平台会如何变化。

And even right now, OpenAI is trying to tap into becoming a new social media platform.

而即便是现在，OpenAI 也在尝试切入、成为一个新的社交媒体平台。

Is there somewhere down the road a possibility that you could also become a social media platform?

那未来某个时点，你们是不是也有可能成为一个社交媒体平台？

No, we don't want to be that.

不，我们不想成为那种东西。

I know you mentioned privately that our statement of building a creative operating system is a bold statement.

我知道你私下提过，我们说要做一个"创意操作系统"（creative operating system）是个挺大胆的说法。

But let me clarify it a little bit.

但我想稍微澄清一下。

When it comes to larger teams — like over 50 people — it typically involves managing several social media channels.

当它面向更大的团队时——比如 50 人以上——通常就涉及管理好几个社交媒体渠道。

And really trying to make sense of content and data.

以及真正去理解内容和数据。

So that's what I mean by a creative operating system: how I can make relevant content today for those different social medias, and how I can make sense of the results, and do better tomorrow.

所以这就是我说的"创意操作系统"的意思：我今天怎么为不同的社交媒体做出对路的内容，又怎么从结果里读出门道，明天做得更好。

So have you already built an engine, or a way, to quickly monitor what's a hit on social media —

那你们是不是已经造了一个引擎，或者一套办法，能快速监测社交媒体上什么内容火了——

and then try to build the content as quickly as possible, and ship it?

然后尽可能快地把内容做出来、发出去？

We're actually working with partners on that, yes.

其实我们正在和合作伙伴一起做这件事，是的。

That's a top priority.

这是头号优先级。

OK, nice, nice. We've got a few more minutes, so let's talk a little about the vision and the future.

好，很好，很好。我们还剩几分钟，可以稍微聊聊愿景和未来。

I think overall you've seen and tested a lot of models — you see the pros and cons of the current ones, and you definitely see they're evolving.

我觉得整体上你看过、也测试过很多模型，你清楚当前这些模型的优劣，也确实看到它们在不断进化。

So, riding with the trend, in two or three years, what are some use cases that you think Higgsfield is going to unlock?

那顺着这股趋势，未来两三年里，你觉得 Higgsfield 会解锁哪些用例？

In three years, we're gonna be at the moment where the impact of video AI and social media content is just gonna be so profound.

三年后，我们会走到这样一个时刻：视频 AI 和社交媒体内容的影响，会变得极其深远。

We're talking about tens of millions — maybe 10 million social media professionals making videos using AI.

我们说的是数千万——也许是 1000 万社交媒体专业人士在用 AI 做视频。

I think ultimately the diversity of content is going to drastically increase.

我觉得最终，内容的多样性会大幅提升。

But also, I think it's going to make things a bit more balanced.

但同时，我觉得它也会让格局变得更均衡一些。

Let me tell you a funny story.

我给你讲个有意思的故事。

One of our key clients uses Higgsfield to make UGC ads for a wellness product, targeting people in their 50s and 60s.

我们的一个核心客户，用 Higgsfield 为一款健康（wellness）产品做 UGC 广告，目标人群是五六十岁的人。

It's almost impossible for them to find a creator on the TikTok marketplace who's gonna represent the brand.

对他们来说，几乎不可能在 TikTok 的达人市场（marketplace）上找到一个愿意代言这个品牌的创作者。

Really an opportunity — a marketing opportunity.

确实是个机会——一个营销机会。

Yes. So, their products.

是的。所以，他们的产品。

Yeah. And they just say — look, all this Barbie/cereal-type of marketing doesn't work for these wellness products.

对。他们就说——你看，这种"芭比／麦片广告"式的营销，对这些健康产品根本不管用。

It kind of needs to show the transformation, right?

它需要的是展示一种"转变"（transformation），对吧？

Market researchers need to show the transformation.

市场研究者得把这种转变呈现出来。

So I think that's just one example.

所以我觉得这只是一个例子。

And today, it's difficult for us to even scope it, because social media — we live in this attention economy — became the most important form of media, and it's gonna be drastically transformed.

而今天，我们甚至很难去界定它的边界，因为社交媒体——我们生活在注意力经济里——已经成了最重要的媒介形态，而它会被彻底重塑。

And I hope that it leads to a higher level of diversity.

我希望这能带来更高水平的多样性。

We saw that slop content, brain rot — like the animated content people are pushing out of Sora — is just one of the examples.

我们看到那种粗制滥造的内容、"脑腐"——就像人们从 Sora 里产出的那些动画内容——只是其中一个例子。

But there's gonna be much more than that.

但远不止于此。

Slop is just the easiest one, that's why we see it first.

Slop 只是最容易做的那种，所以我们最先看到它。

OK. And also, the film industry — the movie industry — do you see some disruption, or it being disrupted, in the next two or three years?

好。还有，电影行业——影视行业——你觉得未来两三年里会出现一些颠覆，或者说它会被颠覆吗？

Or is there still a long time to go?

还是说，还得很久？

I'm not really focused on this market.

我其实不太关注这个市场。

There are so many external forces, and I'm not sure.

外部变量太多了，我也说不准。

Obviously, when you think about some types of videos — say, a Sydney Sweeney ad — it's gonna stay.

显然，当你想到某些类型的视频——比如说一支 Sydney Sweeney 的广告——它会留下来。

Because people want to see Sydney; no AI can make it exactly pixel-perfect.

因为人们想看的就是 Sydney；没有哪个 AI 能把它做得像素级完美。

And this video can be reused many times — it doesn't have the life cycle of just one day.

而且这条视频可以反复使用很多次——它的生命周期不止一天。

Most of social media has a life cycle of one day.

而社交媒体上的大部分内容，生命周期只有一天。

So there's gonna be a place for physical production of genuine, original content — which has a lifespan of more than a day.

所以，真人实拍那些真正原创的内容，依然会有一席之地——它们的寿命超过一天。

But that's gonna be just the premium part of the market, maybe 10 to 20 percent.

但那只会是市场里的高端部分，也许 10% 到 20%。

Some people are gonna keep recording themselves — of course, that'll be a substantial part of the market — but the rest is gonna be gen AI.

有些人会继续拍自己——当然，这仍然会是市场里相当大的一块——但剩下的部分，都会是生成式 AI。

Right. You also talked about the biggest issue in the AI generative video industry being high cost.

对。你也提到，现在 AI 生成视频行业最大的问题是成本高。

And I assume a lot of that cost is just the cost of using the models.

我猜其中很大一部分成本，就是使用模型本身的成本。

In the next three years, is there any product direction you want to optimize, so you can mitigate some of the cost-related issues?

那在接下来三年里，你有没有想优化的某个产品方向，好缓解一些和成本相关的问题？

Of course. As these use cases mature, we're running a lot of post-training and optimization to bring the cost ten times down.

当然。随着这些用例成熟，我们在做大量后训练和优化，把成本降到原来的十分之一。

So that's where we could differentiate.

这正是我们能做出差异化的地方。

For, let's say, 5 to 10 percent of the use cases that drive most of the adoption, we can deliver very efficient models.

对那些驱动了大部分采用的、比如 5% 到 10% 的用例，我们可以提供非常高效的模型。

Gotcha, gotcha. But there'll always be new models coming out — people will always chase the new ones.

懂了，懂了。但总会有新模型出来——人们总会去追最新的。

Yeah, absolutely.

对，完全是。

The next six months, it's likely to be the case, as the pace of innovation is like one new model a week.

未来六个月很可能还是这样，因为创新的节奏差不多是一周一个新模型。

But at some point, these models get bigger, and the pace of innovation is gonna shift to another level — slower.

但到某个时点，这些模型会变得更大，创新的节奏会换到另一个量级——变慢。

Like, when a new DeepSeek comes out, or Kimi, or whatever, it doesn't necessarily make people switch from OpenAI.

就像，当一个新的 DeepSeek、或 Kimi、或别的什么出来时，并不一定会让人们从 OpenAI 切换走。

We're gonna get to that level of maturity with video.

视频也会达到那种成熟度。

Video is probably the most difficult domain — that's why it takes more time for the industry to mature, and it's probably like two years behind the coding space.

视频大概是最难的领域——所以行业需要更长时间才能成熟，它大概比编码领域晚两年。

We're gonna get there as well.

我们最终也会走到那一步。

Yeah, agree, agree.

对，同意，同意。

So maybe one prediction I hope you can make: last week I read an article saying that AI articles now take up over 50 percent of all web article content.

那也许有个预测我希望你来做：上周我读到一篇文章说，现在 AI 写的文章已经占了全网文章内容的 50% 以上。

So in your prediction, how about video content?

那按你的预测，视频内容会怎样？

The same is true — but this half is gonna be generated in the next three years.

视频也一样——只不过这"一半"会在未来三年里被生成出来。

We'll be talking about double-digit percent of content being AI-generated in the next three years, for sure.

未来三年，我们肯定会谈到两位数百分比的内容是 AI 生成的。

I think earlier, but the max timeline is five years.

我觉得会更早，但最长的时间线是五年。

And my bold prediction is: out of the current companies still in the market, there are gonna be losers who cumulatively lose more than half a trillion dollars in market cap.

而我的大胆预测是：在当前还留在市场上的公司里，会有一批输家，它们的市值累计蒸发超过 5000 亿美元。

And there are gonna be some companies that gain this half a trillion in cap.

同时也会有一些公司，赢下这 5000 亿美元的市值。

And that's gonna happen in the next three years.

而这会在未来三年内发生。

In the space of social media video, we just don't know who it is.

在社交媒体视频这个领域，我们只是还不知道是谁。

But I think there are clearly gonna be winners and losers.

但我认为，赢家和输家会泾渭分明。

All right, so I think we're coming to an end.

好，我觉得我们快到尾声了。

We have two more quick-fire questions that we prepared — just say whatever comes to mind first, and answer it.

我们准备了两个快问快答的问题——脑子里第一反应是什么就答什么。

So I have one question: what's the most counterintuitive decision you've made in the past years, after building Higgsfield?

我手上有一个问题：在做 Higgsfield 这些年里，你做过的最反直觉的决定是什么？

Of course. I think that's a very easy question.

当然。我觉得这个问题很好答。

We got a lot of advice from investors that video AI is an extremely — what's it called — red ocean, a very saturated space.

我们从投资人那里听到很多建议，说视频 AI 是个极其——怎么说来着——红海，非常饱和的领域。

A lot of companies, difficult to differentiate, and that we shouldn't be pursuing these consumer markets.

说很多公司难以做出差异化，我们不该去追这些消费级市场。

But look — even today, no one has a viable team plan and storyboarding; those are table stakes, and we're filling this gap with Higgsfield.

但你看——哪怕到今天，也没人做出可用的团队版和故事板；这些本该是"入场门槛"（table stakes），而我们正在用 Higgsfield 填这个空白。

And I think we're just very early.

而且我觉得我们还非常早期。

So naturally, most of these companies — maybe except Google — will have to verticalize over time.

所以很自然地，这些公司里的大多数——也许除了 Google——迟早都得走向垂直化。

So, yeah — really being less receptive to what some investors say, and really focusing on adoption patterns, I think is the most difficult decision.

所以，对——真正少去听某些投资人说什么，真正盯住采用数据（adoption patterns），我觉得是最难的那个决定。

OK. Nice.

好，很棒。

And then, what's your most exciting moment since you started Higgsfield?

那么，自从开始做 Higgsfield 以来，你最激动的时刻是什么？

The most exciting moment for me was to see... oh, but that's a long question.

对我来说，最激动的时刻是看到……哦，不过这是个大问题。

I think it's probably not the right time and place to talk deeper about that.

我觉得现在可能不太适合深聊那个。

But I'm really excited about this new model of AI agencies — like teams of maybe up to 10 or 20 people who are revolutionizing the whole bigger-agency model.

但我真的很为这种新型的 AI 代理公司（agency）感到兴奋——就是那种大概 10 人、20 人以内的小团队，正在彻底改写传统大型代理公司的模式。

In the agency world, I see a lot of consolidation on one side, and it may look like the bigger players only get bigger.

在代理行业里，我看到一边是大量整合，看上去好像只有大玩家会越来越大。

But these new agencies redeveloped the process completely with AI.

但这些新型代理公司，用 AI 把整个流程完全重做了一遍。

So seeing that Higgsfield can drive their success was very important.

所以看到 Higgsfield 能助推他们成功，对我来说非常重要。

And most of these agencies come from emerging markets.

而且这些代理公司大多来自新兴市场。

So seeing that it provides net-new economic opportunities — that's something that excites me.

所以看到它带来了净新增的经济机会——这就是让我兴奋的东西。

Super excited, super excited.

太激动了，太激动了。

All right, I think that's the end.

好，我觉得到这里就结束了。

Thank you, and thank you so much, Alex, for coming today.

谢谢，非常感谢 Alex 今天的到来。

We really hope we can invite you to interview again in the next six months — and we're gonna see a huge change in this market as well.

我们真心希望未来六个月能再请你来聊一次——到时这个市场也会有翻天覆地的变化。

Sounds good.

听起来不错。

Thank you.

谢谢。

I'm so excited to see what comes next. Thank you, thank you very much.

我太期待接下来会发生什么了。谢谢，非常感谢。