Alex Mashrabov, the former head of generative AI at Snap, saw a growing opportunity in the world of personalized video creation — and he ran with it. Enter Higgsfield AI, a new video generation platform aimed at helping creators of all types produce high-quality, customized content with just a few taps.
At the heart of Higgsfield is its first app, Diffuse, powered by a proprietary text-to-video model. Whether you're starting from a selfie or a text prompt, Diffuse can create dynamic, personalized videos — even inserting you directly into the action.
“Our target audience is creators of all kinds,” Mashrabov said in an interview with TechCrunch. “That includes casual users making fun clips with friends, influencers experimenting with new content formats, and social media marketers who want to stand out.”
A Track Record of Innovation
Mashrabov’s journey to Higgsfield started with AI Factory, his previous startup, which was snapped up by Snap in 2020 for a cool $166 million. While at Snap, he played a key role in building some of the platform’s most recognizable AI products — from AR filters to Snapchat Cameos, and even the infamous MyAI chatbot.
Now, with co-founder Yerzat Dulat, an AI researcher with deep expertise in generative video, Mashrabov is turning his sights toward consumer-friendly AI tools that offer a surprising amount of control.
What Makes Diffuse Different?
With Diffuse, you can:
- Browse a curated selection of AI-generated video clips
- Upload your own reference media (images or videos)
- Use prompts to describe characters, actions, and scenes
- Drop your digital likeness into a scene — complete with expressive facial movements and lifelike gestures
“We’re building what we call ‘world models’ for consumers,” Mashrabov explained. “This lets us offer top-tier video generation and editing with intuitive, fine-grained control.”
But Higgsfield isn’t the only name in the game. Industry heavyweights like Runway and Haiper are also pushing the limits of generative video. So what gives Higgsfield an edge?
Mobile-First, Social-Ready
Higgsfield is betting big on a mobile-first approach, focusing on iOS and Android rather than traditional desktop workflows. The strategy is clear: creators want tools they can use on the go, with minimal friction.
“By starting with mobile, we’re able to prioritize ease of use and build for social content creation from day one,” Mashrabov said.
Despite the sophistication of the tech, Higgsfield is running lean. A 16-person team built the core generative model in just nine months, training it on a cluster of 32 GPUs. (For comparison, OpenAI reportedly uses tens of thousands.)
To date, the startup has raised $8 million, mostly from a recent seed round led by Menlo Ventures. The funding will go toward building out an enhanced video editor and training even more advanced models — specifically tuned for social media applications.
Monetization, Moderation, and the Road Ahead
Right now, Diffuse is free to use, but the long-term plan includes a monetization model targeting marketers — think subscriptions or campaign-based pricing for premium features.
“We’re hearing from CMOs and creative leads that they’re under pressure to create more content, faster, and with tighter budgets,” Mashrabov said. “Generative video tools like Diffuse are perfectly positioned to help with that.”
Still, Higgsfield faces some of the same challenges as other AI startups. Like many generative models, theirs could potentially replicate copyrighted material if not properly trained. Mashrabov stayed tight-lipped about the training data, only saying it came from “multiple publicly available” sources. He also didn’t comment on whether user data might be used to train future models — though users can request to have their data deleted at any time.
And, of course, with powerful personalization tools comes the potential for misuse. Think deepfakes, content theft, and other forms of digital manipulation. While Mashrabov didn’t go into detail, he said Higgsfield uses a mix of automated and manual moderation and is rolling out the product gradually to help monitor and mitigate abuse.
Final Thoughts
Higgsfield AI is setting out to redefine how creators interact with video — by putting generative tools directly into their hands, wherever they are. The competition is fierce, and the challenges are real, but with a strong team, a bold vision, and a mobile-first mindset, Mashrabov and his team are betting that personalized AI video creation is the future of content.
Validate your login
Sign In
Create New Account