❶

Kling AI Try On

Now here is an AI tool that (if widely adopted) actually might be worth highlighting as a reason for that expected e-commerce boom a lot of people have been forecasting.

Here are the details:

AI-Powered Try-On: Generates realistic images of users wearing selected clothing, adapting to body type and pose.

Customization: Upload your photo and browse virtual clothing options.

Realistic Results: Advanced AI is supposed to accurately shows how clothes drape and fit.

Simple Process: Just upload a photo and pick clothes to see instant visualizations.

Pretty slick! We’ve heard of this before—both Amazon & Walmart have been working on/teasing similar features earlier this year, albeit limited to certain products. So it’s really the realism and capacity for showcasing realistic “fit” for clothes that has people buzzing about this new drop from Kling. Also the fact it seems to be actually pretty easy to use.

However, like most things in the AI space, its application will probably come down to cost. If you’ve played around with video generators, you know they aren’t cheap.

For example:

The Runway ML standard plan costs $12 per month and provides 625 credits. Each video generation using Gen-3 Alpha requires 10 credits per second of video. For a 10-second video, which is one of the available export durations, we need 100 credits. To calculate the average cost per video generation:

Average Cost=Monthly Membership Cost Monthly Credits Credits per Video Average Cost= Monthly Credits Credits per Video Monthly Membership Cost

Average Cost=$12(625/100)=$1.92Average Cost=(625/100)$12=$1.92

Therefore, the average cost for generating a 10-second video using Runway ML's Standard plan is $1.92.

I’m not saying it’s going to be exactly equivalent with try-on videos, but just think about that for a moment. Who bears the cost? Consumer or vendor?

Some high end vendors might provide virtual try on for free…but if I was a betting man I’d bet pretty heavily on groups of internet trolls running try on videos non-stop just to screw the company before long.

So it’s probably going to require a premium subscription for shoppers, and come with a limit at that. For example, Amazon Prime members enjoy 3 free try-ons per week with their Prime membership.

Either way, one could also wonder if part of the magic of shopping is dispelled by actually seeing the item on yourself first. We are usually buying the dream, right? Despite the rise of inclusive modeling, the more base level of consumer psychology implies that people purchase things that they feel will make them better somehow, or more similar to the people they see wearing them (the models included).

So perhaps seeing oneself wearing the item first could actually end up doing more to restrict purchases, than so fuel them.

In any case, fun thing to play with for us for now, and probably just a cute tool for premium shoppers in the future.

❷

AI Agent State of Affairs

Much has been made of AI Agents and their incoming effects on software, but separating hype from reality isn’t easy as always.

And when the hype is something as far-out as “agents can automate literally every task for you” it’s pretty hard to put a finger on what’s actually useful now. This thread does a great job of highlighting the ways people are building useful agents at this very moment.

The key piece for understanding the hype is here:

Like most things in AI, the underlying potential seems to rely on the idea of what could happen when AI development goes so far as to make building tech products available to non-tech people.

“Anyone can produce a Hollywood movie, you just need an idea”

“Anyone can make Taylor Swift-quality music [which isn’t that great imo 🤮], you just need an idea”

“Anyone can write software, you just need an idea”

It is pretty cool to imagine that kind of world. But we are probably vastly overestimating the amount of time it will take to get to that kind of reality. Sure, everyone seems to have a smartphone now, but ask them to reset the cell tower settings, we probably lose about 90% of the population.

What is clear and present though, is that new possibilities are wide open for those that are able to quickly master these new tools. Imagining the different things that can be built within what our tech is already capable of providing is the first step. Sending a piece of text information from one phone to another was pretty cool. Sending a picture through that same channel, even cooler. Sending series of 24 pictures laced together in each second that play as if real life (videos)? Insane.

But that wasn’t really original, was it?

We were sending text written on pieces of paper through the mail for 100 years before that.

On pieces of parchment delivered by pony riders before that.

And verbal messages conveyed by marathon runners even before that.

So perhaps we should think first about what people will ultimately want to do on a smartphone that talks to you like ChatGPT. It’s probably pretty much what we do already, just done a lot faster.

❸

Flux LORA Consistent Character Training

This user shares a nice workflow for generating custom characters with consistency. But at this point, do we even care about custom workflows? We should be expecting most AI video tools to offer this level of consistency. Here are a few that are already announcing similar features within existing toolkits.

• KLING: developed by Chinese company Kuaishou, has introduced a groundbreaking feature called "Custom Models" that significantly improves character consistency in generated videos.

• Leonardo's Character Reference: This tool enables creators to generate consistent characters across various images and scenes, simplifying workflows in film, TV, fashion, marketing, and storytelling.

• Luma's Dream Machine: Available on web and iOS, Dream Machine allows users to create videos with consistent characters by understanding interactions between people, animals, and objects within the physical world.

• RenderNet AI: This platform provides advanced control over character design, composition, and style, enabling the generation of consistent characters in both images and videos.

• Atlabs: Atlabs offers AI-driven video creation tools that ensure character consistency, enhancing the depth and realism of storytelling across various media formats.

Consistent character generation is one of those things creatives have seemed to be waiting on before really diving into using AI tools. It’s crucial for making AI usable for content creators and creatives for several key reasons:

Storytelling Coherence

When characters maintain consistent appearance and traits across scenes, audiences can:

• Form stronger emotional connections

• Follow stories without distractions

• Become fully immersed in the world

It’s absolutely essential for long-form content like comics and animated series.

Brand Identity and Recognition

For brand and franchise development:

• Creates recognizable visual identity

• Builds brand loyalty

• Enables successful merchandising

Production Efficiency

AI-generated consistent characters streamline creation by:

• Reducing manual adjustments

• Saving time and resources

• Allowing focus on creative aspects

Collaborative Workflows

Benefits for team environments:

• Provides unified reference point

• Maintains consistency across team members

• Simplifies style guide creation

Adaptability and Scalability

Offers flexibility through:

• Easy character adaptation

• Diverse content creation

• Quality-maintained scaling

Enhanced Creative Control

Gives creators better control with:

• Fine-tuning capabilities

• Character variation options

• Cross-style consistency

Consistent character generation is definitely a game-changer for AI content creation. It makes storytelling smoother, helps build brands, and speeds up production. And it's getting better fast. The creators who jump on this tech early will have a serious edge in pumping out quality content at scale.