Higgsfield AI is a mobile-first artificial intelligence platform developed to produce image-based personalized video content. Founded in 2024 by Alex Mashrabov and Yerzat Dulat, the platform focuses on AI models designed for generating short, social media-oriented videos. The company is headquartered in the United States.
Founders and Background
Alex Mashrabov, one of the co-founders of Higgsfield AI, previously led the generative AI division at Snap Inc. He was also a co-founder of AI Factory, a startup later acquired by Snap. Yerzat Dulat, the other co-founder, is an AI researcher with a background in generative video technologies. The founders’ prior experience has contributed to the platform’s focus on social media content creation and mobile user experience.
Diffuse Application
Diffuse, the first product launched by Higgsfield AI, was introduced in a preview version in 2024. It is an image-to-video transformation application developed for smartphones. The application enables users to generate personalized, character-focused videos of two seconds in length using selfie images.
Features
- Users can create personalized videos by selecting from video templates within the app or by defining scenes from scratch using text, images, or video inputs.
- A Prompt Builder tool allows scene generation based on textual and media prompts.
- The app includes animation of facial expressions and gestures.
- Video duration is limited to 2 seconds in the preview version.
- As of 2024, the app has been gradually released in India, South Africa, the Philippines, Canada, and Central Asia. It is available on iOS, with an Android version under development.
Technological Infrastructure and Model Development
Higgsfield AI is developing a custom-built video-oriented AI model. The company’s video generation system is based on transformer architecture, similar to that used in large language models. The model has been trained on a limited number of GPUs (32), utilizing a proprietary training framework developed by the company. The core objective of the model is to generate realistic, detailed, and fluid video content directly on mobile devices.
Cinematic Camera Motion and Narrative Structure
As of 2025, Higgsfield AI has expanded its focus to include not only visual accuracy but also cinematic language and narrative structure. A control engine developed for this purpose enables users to replicate the following cinematic camera movements using a single image and a basic text input:
- Dolly-in and crash zoom
- Bird’s-eye view
- Body-mounted rig
This system simulates traditional camera motions typically achieved through professional film equipment. As a result, individual creators and small production teams can incorporate advanced cinematic techniques into their video projects using AI.
Use Cases
Higgsfield AI is used by individual users, social media content creators, and digital marketing professionals to generate short, personalized videos. The platform’s target audience includes broad user groups seeking fast and creative video generation via mobile devices.
Sample Video (Source: Higgsfield AI)
Security and Data Policies
The platform states that it has implemented certain safeguards for user data. Users may request the deletion of their generated content. Both automated and manual review systems are employed within the Diffuse app to prevent misuse. However, the source of training data and whether user-generated content is used for model training remain unspecified.
Funding and Business Model
Higgsfield AI has received over $8 million in seed funding, with a significant portion provided by Menlo Ventures. The platform is currently available for free, while high-volume content production is subject to subscription-based pricing plans.
Competition and Market Position
Higgsfield AI operates in the same market as platforms such as Runway, Pika Labs, Haiper, and OpenAI’s video generation system, Sora. Its differentiating aspects include a mobile-first design approach, personalized character generation, and cinematic camera language capabilities.
