TENCENT HUNYUANPORTRAIT, an open source AI model for drawings

Tuesday has released a new artificial intelligence model (AI) on Tuesday can move pictures that are still a picture. The Great Language model (LLM) is based on Hunyuanportrait, and it depends on the structure of spread, and it can create video clips with realistic animation based on a reference image and guiding video. The researchers behind the project highlighted that the model can capture both facial data and spatial movements to accurately sync it in the reference image. Tincent now opens the Hunyuanportrait AI model, can be downloaded and operated locally from the famous warehouses.

Tencent’s Hunyuanportrait can bring pictures that still be to life

in mail On X (previously known as Twitter), the official Testent Hunyuan’s official handle announced that the Hunyuanportrait model is now available for the open community. The artificial intelligence model can be downloaded from Tencent’s Gyrroup and Embroidery Lists. In addition, before printing paper Form details are also hosted on Arxiv. It is worth noting that the artificial intelligence model is available for academic and research -based use situations, but not for commercial use.

Hunyuanportrait can create vibrant animated videos using a reference image and driving videos. It takes the face data, puts the head from the video, and turns it to the image of the image that is still. The company claims that the movement synchronization is accurate, and even micro -face expression changes are repeated.

Honoyuanuratrit architecture
Credit image

On its model pageThe researchers in Tennetnet detailed the structure of Hunyuanportrait. It is based on the structure of stable proliferation models, as well as the core control of the situation. These pre -trained encoders separate movement and identity information in videos. Data is taken as control signals, which are then injected into the fixed image via UNET The company claims that this brings both spatial accuracy as well as time consistency in the output.

Tencent claims that the artificial intelligence model outperforms existing alternatives that are open source over the parameters of temporal consistency and control, but these scales have not been independently verified.

These models can be useful in the film and animation industry industries. Traditionally, manually manually facial expressions or use expensive motion capture systems to move the characters realistically. They will allow models such as Hunyuanportrait to feed the design of letters, targeted movements and facial expressions, and will be able to generate the output. This LLMS also has the ability to make high -quality animation available for smaller studios and independent creators.