Today's Image & Video Generation: Fastest-Growing Projects — April 29, 2026
This week, the Image & Video Generation space on GitHub has seen a surge in activity around text-to-image and image editing tools. The trend is clear: developers are hungry for more efficient and user-friendly ways to generate high-quality images and videos using AI. With growth scores soaring across the board, it's an exciting time to explore what's new and next in this rapidly evolving field.
First up is lidge-jun/ima2-gen, a minimal CLI + web UI for OpenAI GPT Image 2 generation that boasts a whopping Growth Score of 48.75 and 117 stars. Its popularity stems from its dual authentication options - API Key (paid) or OAuth via ChatGPT (free) - making it an attractive choice for developers seeking flexibility in their image generation workflows.
Next, jd-opensource/JoyAI-Image has captured the attention of over 1,966 star-gazers with its unified multimodal foundation model for image understanding, text-to-image generation, and instruction-guided image editing. With a respectable Growth Score of 44.81, JoyAI-Image is rapidly becoming a go-to solution for those seeking to tap into the power of multimodal AI.
Baidu's ERNIE-Image has garnered significant interest with its open text-to-image generation model built on a single-stream Diffusion Transformer (DiT). Sporting a Growth Score of 19.90 and 396 stars, ERNIE-Image is winning fans for its impressive performance among open-weight text-to-image models.
In the realm of video generation, OpenImagingLab/AnyRecon has made waves with its Arbitrary-View 3D Reconstruction using Video Diffusion Model. With a Growth Score of 16.10 and 217 stars, AnyRecon is proving to be an essential tool for developers working on cutting-edge computer vision projects.
OpenMOSS/MOSS-VL rounds out the top five with its core multimodal model series dedicated to visual understanding. Its respectable Growth Score of 12.81 and 238 stars speak to the growing interest in MOSS-VL as a versatile solution for image and video generation tasks.
Further down the list, notable mentions include jiangmuran/claude-image (Growth Score: 11.83, Stars: 29) which teaches agents to use GPT Image 2 effectively; gulucaptain/Camera-Transformer-1 (Growth Score: 11.37, Stars: 222) which generates videos with precise camera control; and ShandaAI/AlayaRenderer (Growth Score: 10.89, Stars: 531) an AI-native Renderer for Games and Virtual Worlds.
Other projects worth keeping an eye on include kangarooking/design-image-studio (Growth Score: 9.41, Stars: 87), which generates high-quality designs from fuzzy visual demands; and omni2sound/Omni2Sound (Growth Score: 6.17, Stars: 33) a multimodal audio generation codebase.
While some tools like OpenImagingLab/AnyRecon are pushing the boundaries of what's possible with video diffusion models, others like lidge-jun/ima2-gen and jd-opensource/JoyAI-Image are focusing on making text-to-image and image editing more accessible. One thing is clear: this space will continue to evolve at breakneck speed as innovators keep pushing the limits of AI-powered creativity.
First up is lidge-jun/ima2-gen, a minimal CLI + web UI for OpenAI GPT Image 2 generation that boasts a whopping Growth Score of 48.75 and 117 stars. Its popularity stems from its dual authentication options - API Key (paid) or OAuth via ChatGPT (free) - making it an attractive choice for developers seeking flexibility in their image generation workflows.
Next, jd-opensource/JoyAI-Image has captured the attention of over 1,966 star-gazers with its unified multimodal foundation model for image understanding, text-to-image generation, and instruction-guided image editing. With a respectable Growth Score of 44.81, JoyAI-Image is rapidly becoming a go-to solution for those seeking to tap into the power of multimodal AI.
Baidu's ERNIE-Image has garnered significant interest with its open text-to-image generation model built on a single-stream Diffusion Transformer (DiT). Sporting a Growth Score of 19.90 and 396 stars, ERNIE-Image is winning fans for its impressive performance among open-weight text-to-image models.
In the realm of video generation, OpenImagingLab/AnyRecon has made waves with its Arbitrary-View 3D Reconstruction using Video Diffusion Model. With a Growth Score of 16.10 and 217 stars, AnyRecon is proving to be an essential tool for developers working on cutting-edge computer vision projects.
OpenMOSS/MOSS-VL rounds out the top five with its core multimodal model series dedicated to visual understanding. Its respectable Growth Score of 12.81 and 238 stars speak to the growing interest in MOSS-VL as a versatile solution for image and video generation tasks.
Further down the list, notable mentions include jiangmuran/claude-image (Growth Score: 11.83, Stars: 29) which teaches agents to use GPT Image 2 effectively; gulucaptain/Camera-Transformer-1 (Growth Score: 11.37, Stars: 222) which generates videos with precise camera control; and ShandaAI/AlayaRenderer (Growth Score: 10.89, Stars: 531) an AI-native Renderer for Games and Virtual Worlds.
Other projects worth keeping an eye on include kangarooking/design-image-studio (Growth Score: 9.41, Stars: 87), which generates high-quality designs from fuzzy visual demands; and omni2sound/Omni2Sound (Growth Score: 6.17, Stars: 33) a multimodal audio generation codebase.
While some tools like OpenImagingLab/AnyRecon are pushing the boundaries of what's possible with video diffusion models, others like lidge-jun/ima2-gen and jd-opensource/JoyAI-Image are focusing on making text-to-image and image editing more accessible. One thing is clear: this space will continue to evolve at breakneck speed as innovators keep pushing the limits of AI-powered creativity.