AI Video Startup Raises $8 Million to Animate Social Media Characters

San Francisco-based AI developer Higgsfield announced a successful $8 million seed fundraising round on Wednesday, funds that the firm says it will pour back into its video creation tools for marketers creating content for social media.

The company says its first priority will be the responsible development and democratization of AI video creation.

Higgsfield is focused on supporting social media content creators, co-founder and CEO Alex Mashrabov told Decrypt, not to create feature-length AI-generated videos like some of its competitors.

“While [OpenAI’s] Sora is compelling, we don't have a goal to generate two-hour-long videos for like movies,” Mashrabov said. “Our goal is to empower marketing departments for social media marketing.”

Mashrabov is the former director of generative AI at social media giant Snap, which acquired his startup AI Factory in 2018. He co-founded Higgsfield only last year.

“Higgsfield is my third startup and I have an engineering background,” Mashrabov said. “In the first startup, we did an application which helped teenagers solve math problems by taking a picture of it.”

The $8 million raise was led by Menlo Ventures with participation from Charge Ventures, Bitkraft Ventures, K5 Tokyo Black, DVC, and AI Capital Partners, the U.S.-based fund from Luxembourg-based Alpha Intelligence Capital.

Others joining the funding round include angel investors Jack Brody, Nima Khajehnouri, Ilya Gelfenbeyn, Abhay Parasnis, and Amar Hanspal.

Higgsfield’s flagship mobile application, Diffuse, creates animated videos based on user prompts. With it, users can create AI-generated clips based on photos they upload to the app. Diffuse is available on Android and iOS platforms and touts ease of use, without requiring deep technical knowledge.

The new funding will go towards further refining their mobile AI video model, Mashrabov said.

“We made [Diffuse] on mobile because we want to start learning from users, and we are already doing that,” Mashrabov said. “We want to get users tools in their hands, and we know a lot of social media marketing—especially entry-level, freelance social media marketing—and conversations happen on the mobile phone.”

As Mashrabov explained, the “cartoon” animation style in Diffuse is not just about being creative. It's a way to keep the application from being used to create AI-generated deepfakes.

“If you look at the app, most of the videos are sort of cartoony, and we made it that way on purpose,” Mashrabov said. “Because the cartoony type of content doesn't pretend to be realistic, and it's sort of a stepping stone for us before we introduce a more powerful and more realistic model.”

Mashrabov said there is a need for quick content production in social media, as long production timelines are impractical.

“Pretty much what social media requires from people is to stay on top of trends,” he said. “When someone has a content production pipeline of three weeks and outsources it somewhere else in the world, then it doesn't work.”

There has also been a shift from desktop to mobile video editing driven by tools like CapCut and Canva. But there remain inefficiencies in video production, and that's what Higgsfield will tackle.

Diffuse has already been launched in select countries, with more markets coming soon, according to Mashrabov.

“We picked South Africa, Canada, India, and the Philippines because these are the markets that have an English-speaking population,” Mashrabov said. "This is our priority, to learn from them, number one, and number two is making sure that the videos are not going to be misused.”

AI-generated deepfakes remain a hot-button issue given the proliferation of AI tools since the launch of GPT-4 last year. Deepfakes of world leaders, including Pope Francis, U.S. President Joe Biden, and former President Donald Trump, have gone viral online.

Last month, the Biden Administration announced plans to use cryptography and watermarking on White House communications to stop deepfakes and misinformation.

Higgsfield is also working with industry leaders to add AI watermarking to the platform—but Mashrabov conceded that the participation of other major content and social media platforms is needed to be able to effectively authenticate and attribute content.

“All these platforms are developing their AI policies, and I think definitely there is a goal to achieve transparency that the content is AI generated and making sure we are complying with approaches taken by major social media platforms.”

Edited by Ryan Ozawa.