Higgsfield AI, a video AI firm, has launched its first synthetic intelligence (AI)-powered app for smartphones. Dubbed Diffuse, this mobile-first app is an image-to-video generator that may use a selfie to show it right into a video that includes the character. The corporate claims that the character it creates comes with lifelike movement. The app is being step by step rolled out to pick markets and customers will be capable to discover the app on each Android and iOS. This AI instrument will probably compete with OpenAI’s Sora, which is reported to be launched later this 12 months.
The corporate’s official account took to X (previously often known as Twitter) to make the announcement. Higgsfield AI launched itself as a video AI firm which is concentrated on creating video content material for social media. This imaginative and prescient probably comes from the CEO of the AI agency, Alex Mashrabov, who was beforehand main the generative AI division at Snap. As per the put up, the corporate is constructing a foundational mannequin from scratch for its instruments which will likely be totally video-focused.
We’re additionally introducing Diffuse – a enjoyable option to attempt a preview of our video mannequin on cell.
Select a video from the content material library, choose your selfie and Diffuse will generate a customized character in that video’s model.
Or use Immediate Builder to create a video from… pic.twitter.com/lN7azXycsN
— Higgsfield AI (@higgsfield_ai) April 3, 2024
In a separate post, the corporate additionally launched Diffuse, the primary cell app constructed by the corporate, which is at present out there in preview mode. “Select a video from the content material library, choose your selfie and Diffuse will generate a customized character in that video’s model. Or use Immediate Builder to create a video from scratch utilizing textual content, photos or video. Diffuse presents a deep stage of personalization, inventive management and fine-tuning so anybody can create precisely what they need (with safeguards after all),” the put up added. The app is at present out there in choose markets which embrace India, South Africa, the Philippines, Canada, and international locations in Central Asia.
We had been capable of find the app within the App Retailer, but it surely didn’t present up on the Android Play Retailer. This will change within the subsequent few days as the corporate is step by step rolling out the app. The app at present presents video technology of two seconds as it’s in preview mode. The AI agency says its final purpose is to realize reasonable, detailed and fluid video, all on a cell machine.
For this, it’s constructing its foundational mannequin from scratch. The video mannequin makes use of transformer architectures, that are additionally utilized by OpenAI’s ChatGPT. Higgsfield AI additional highlights that it was capable of effectively prepare its AI mannequin on restricted GPUs on account of a proprietary framework developed in-house. The corporate has not revealed when it’d launch the total model of the app in public.
