Today's Image & Video Generation: Fastest-Growing Projects — October 22, 2026
This week, the Image & Video Generation space continues to see significant innovation and interest, particularly around open-source projects that offer novel approaches to image super-resolution and pixel-space generation. One standout project is Pixel Loom, an AI-driven editor for creating pixel art and animations with a high growth score, indicating growing community engagement.
Pixel Loom (yangsir2077/pixel-loom) offers an AI-powered tool for generating and editing pixel art and animations on Windows platforms. With 17.07 in growth score and 77 stars, Pixel Loom is gaining traction among developers interested in leveraging AI to enhance their creative processes.
BG2X (isixe/bg2x) is a privacy-first image super-resolution tool that allows for upscaling images without compromising user data privacy. Its high growth score of 14.46 and significant number of commits over the past month suggest active development and increasing interest from users looking to upscale images while maintaining control over their personal information.
Iris-3B (speridlabs/iris-3b) is a diffusion transformer model capable of generating pixel-space images directly, without using VAE. The same model can also perform monocular depth estimation and image restoration. With 200 stars and a growth score of 9.76, Iris-3B has garnered considerable attention for its versatility in handling various vision tasks.
Video Diffusion Avatars (SimonGiebenhain/video-diffusion-avatars) is an official code repository for converting video diffusion models into controllable 3D avatars. Its lower growth score of 6.92 and 47 stars indicate steady interest from researchers and developers in exploring the potential of video diffusion models beyond traditional applications.
JEV_sees (CharlesFeng0314/JEV_sees) provides real-time visual decisions from RGB, video, and RGB-D cameras, enabling various vision-based tasks such as object detection and tracking. With 6.09 growth score and 115 stars, JEV_sees is gaining traction among developers working on computer vision projects that require accurate and efficient processing of camera inputs.
Harshpreet Hand (harshpreet931/harshpreet-hand) showcases an innovative approach to converting handwritten notes into digital fonts using Claude Code. Although its growth score is 5.40 with only 45 stars, the project highlights a unique application of AI in digitizing handwriting and offers insights into the potential use cases for this technology.
HarnessIR (PolyU-VCLab/HarnessIR) aims to improve image restoration by harnessing multimodal foundation models. With a growth score of 4.00 and 34 stars, HarnessIR is attracting interest from researchers interested in leveraging advanced AI techniques to enhance image quality across various real-world scenarios.
InfoVis (VectorSpaceLab/InfoVis) introduces a single model for generating and editing slide images with VisSlideBench as the benchmark. Its lower growth score of 3.31 but steady number of commits suggests ongoing development and interest from researchers interested in visual content creation and manipulation at scale.
Wan-3.0 Video API (ejsyk3xdp3t2o/wan3.0video-wan-3.0-video-api) provides a gateway for video generation with pricing details and examples in both curl and Python formats, offering an OpenAI-compatible interface. With 93 stars and a growth score of 3.16, Wan-3.0 Video API is attracting developers seeking flexible and cost-effective solutions for generating high-quality videos programmatically.
Orbia (worldbench/Orbia) explores whether generated videos form a consistent 3D world through its research project. Its lower growth score of 3.00 with 48 stars indicates steady interest from the community in understanding the limitations and potential of video generation models to create coherent virtual environments.
Today's trends highlight a diverse range of projects that span creative tools, privacy-focused utilities, versatile AI models, and innovative research efforts, each contributing uniquely to the evolving landscape of image and video generation technology.
Pixel Loom (yangsir2077/pixel-loom) offers an AI-powered tool for generating and editing pixel art and animations on Windows platforms. With 17.07 in growth score and 77 stars, Pixel Loom is gaining traction among developers interested in leveraging AI to enhance their creative processes.
BG2X (isixe/bg2x) is a privacy-first image super-resolution tool that allows for upscaling images without compromising user data privacy. Its high growth score of 14.46 and significant number of commits over the past month suggest active development and increasing interest from users looking to upscale images while maintaining control over their personal information.
Iris-3B (speridlabs/iris-3b) is a diffusion transformer model capable of generating pixel-space images directly, without using VAE. The same model can also perform monocular depth estimation and image restoration. With 200 stars and a growth score of 9.76, Iris-3B has garnered considerable attention for its versatility in handling various vision tasks.
Video Diffusion Avatars (SimonGiebenhain/video-diffusion-avatars) is an official code repository for converting video diffusion models into controllable 3D avatars. Its lower growth score of 6.92 and 47 stars indicate steady interest from researchers and developers in exploring the potential of video diffusion models beyond traditional applications.
JEV_sees (CharlesFeng0314/JEV_sees) provides real-time visual decisions from RGB, video, and RGB-D cameras, enabling various vision-based tasks such as object detection and tracking. With 6.09 growth score and 115 stars, JEV_sees is gaining traction among developers working on computer vision projects that require accurate and efficient processing of camera inputs.
Harshpreet Hand (harshpreet931/harshpreet-hand) showcases an innovative approach to converting handwritten notes into digital fonts using Claude Code. Although its growth score is 5.40 with only 45 stars, the project highlights a unique application of AI in digitizing handwriting and offers insights into the potential use cases for this technology.
HarnessIR (PolyU-VCLab/HarnessIR) aims to improve image restoration by harnessing multimodal foundation models. With a growth score of 4.00 and 34 stars, HarnessIR is attracting interest from researchers interested in leveraging advanced AI techniques to enhance image quality across various real-world scenarios.
InfoVis (VectorSpaceLab/InfoVis) introduces a single model for generating and editing slide images with VisSlideBench as the benchmark. Its lower growth score of 3.31 but steady number of commits suggests ongoing development and interest from researchers interested in visual content creation and manipulation at scale.
Wan-3.0 Video API (ejsyk3xdp3t2o/wan3.0video-wan-3.0-video-api) provides a gateway for video generation with pricing details and examples in both curl and Python formats, offering an OpenAI-compatible interface. With 93 stars and a growth score of 3.16, Wan-3.0 Video API is attracting developers seeking flexible and cost-effective solutions for generating high-quality videos programmatically.
Orbia (worldbench/Orbia) explores whether generated videos form a consistent 3D world through its research project. Its lower growth score of 3.00 with 48 stars indicates steady interest from the community in understanding the limitations and potential of video generation models to create coherent virtual environments.
Today's trends highlight a diverse range of projects that span creative tools, privacy-focused utilities, versatile AI models, and innovative research efforts, each contributing uniquely to the evolving landscape of image and video generation technology.