Today's Image & Video Generation: Fastest-Growing Projects — April 28, 2026
This week, the Image & Video Generation space has seen significant activity around text-to-image generation models, with several repositories showcasing innovative approaches to image understanding and editing. Meanwhile, advancements in multimodal foundation models have also caught attention, highlighting the growing interest in unified models for various tasks. As we dive into Today's top-growing repositories, it's clear that researchers are pushing the boundaries of what is possible with AI-generated visuals.
Lidge-jun's ima2-gen repository has seen a remarkable growth score of 55.29 and gained 111 stars, thanks to its minimal CLI + web UI for OpenAI GPT Image 2 generation, offering users a flexible way to generate images with text-to-image, image-to-image, and parallel gen capabilities. The dual authentication system, allowing both paid API Key and free OAuth via ChatGPT, likely contributes to its growing popularity.
With a growth score of 46.30 and an impressive 1,960 stars, jd-opensource's JoyAI-Image repository stands out for its unified multimodal foundation model, which enables image understanding, text-to-image generation, and instruction-guided image editing. The model's comprehensive capabilities make it an attractive choice for researchers and developers seeking a versatile solution.
Baidu's ERNIE-Image repository boasts a growth score of 21.29 and 392 stars, thanks to its open text-to-image generation model built on a single-stream Diffusion Transformer (DiT). With only 8B DiT parameters, it achieves state-of-the-art performance among open-weight text-to-image models, making it an appealing option for those seeking efficient yet high-quality image generation.
The OpenImagingLab's AnyRecon repository has seen moderate growth with a score of 16.17 and 204 stars, likely due to its innovative approach to arbitrary-view 3D reconstruction using video diffusion models. Researchers interested in reconstructing 3D scenes from various viewpoints may find this repository particularly useful.
OpenMOSS's MOSS-VL repository boasts a growth score of 13.40 and 236 stars, thanks to its core multimodal model series dedicated to visual understanding within the OpenMOSS ecosystem. As researchers continue to explore unified models for multiple tasks, MOSS-VL is likely to attract attention from those seeking comprehensive visual understanding solutions.
ShandaAI's AlayaRenderer repository has gained a growth score of 11.29 and an impressive 530 stars, thanks to its AI-native renderer designed specifically for games and virtual worlds. Developers interested in creating immersive experiences may find this renderer particularly appealing due to its ability to generate high-quality visuals.
Gulucaptain's Camera-Transformer-1 repository boasts a growth score of 11.06 and 206 stars, thanks to its innovative approach to video generation with precise, spatially-aware camera control. Researchers interested in generating videos with intentional camera movements may find this repository particularly useful.
Kangarooking's design-image-studio repository has seen moderate growth with a score of 10.40 and 88 stars, likely due to its ability to transform vague visual demands into high-quality, actionable design images. Designers and researchers seeking to automate the design process may find this tool appealing.
NVlabs' PixelDiT repository boasts a growth score of 6.47 and 300 stars, thanks to its innovative approach to image generation using pixel diffusion transformers. Researchers interested in exploring new methods for image synthesis may find this repository particularly useful.
Lastly, YouMind-OpenLab's awesome-gpt-image-2 repository has seen significant attention with a growth score of 5.02 and an impressive 3,423 stars, thanks to its curated library of GPT Image 2 prompts covering various languages and use cases. Researchers seeking to explore the capabilities of OpenAI's next-gen image model may find this resource invaluable.
Lidge-jun's ima2-gen repository has seen a remarkable growth score of 55.29 and gained 111 stars, thanks to its minimal CLI + web UI for OpenAI GPT Image 2 generation, offering users a flexible way to generate images with text-to-image, image-to-image, and parallel gen capabilities. The dual authentication system, allowing both paid API Key and free OAuth via ChatGPT, likely contributes to its growing popularity.
With a growth score of 46.30 and an impressive 1,960 stars, jd-opensource's JoyAI-Image repository stands out for its unified multimodal foundation model, which enables image understanding, text-to-image generation, and instruction-guided image editing. The model's comprehensive capabilities make it an attractive choice for researchers and developers seeking a versatile solution.
Baidu's ERNIE-Image repository boasts a growth score of 21.29 and 392 stars, thanks to its open text-to-image generation model built on a single-stream Diffusion Transformer (DiT). With only 8B DiT parameters, it achieves state-of-the-art performance among open-weight text-to-image models, making it an appealing option for those seeking efficient yet high-quality image generation.
The OpenImagingLab's AnyRecon repository has seen moderate growth with a score of 16.17 and 204 stars, likely due to its innovative approach to arbitrary-view 3D reconstruction using video diffusion models. Researchers interested in reconstructing 3D scenes from various viewpoints may find this repository particularly useful.
OpenMOSS's MOSS-VL repository boasts a growth score of 13.40 and 236 stars, thanks to its core multimodal model series dedicated to visual understanding within the OpenMOSS ecosystem. As researchers continue to explore unified models for multiple tasks, MOSS-VL is likely to attract attention from those seeking comprehensive visual understanding solutions.
ShandaAI's AlayaRenderer repository has gained a growth score of 11.29 and an impressive 530 stars, thanks to its AI-native renderer designed specifically for games and virtual worlds. Developers interested in creating immersive experiences may find this renderer particularly appealing due to its ability to generate high-quality visuals.
Gulucaptain's Camera-Transformer-1 repository boasts a growth score of 11.06 and 206 stars, thanks to its innovative approach to video generation with precise, spatially-aware camera control. Researchers interested in generating videos with intentional camera movements may find this repository particularly useful.
Kangarooking's design-image-studio repository has seen moderate growth with a score of 10.40 and 88 stars, likely due to its ability to transform vague visual demands into high-quality, actionable design images. Designers and researchers seeking to automate the design process may find this tool appealing.
NVlabs' PixelDiT repository boasts a growth score of 6.47 and 300 stars, thanks to its innovative approach to image generation using pixel diffusion transformers. Researchers interested in exploring new methods for image synthesis may find this repository particularly useful.
Lastly, YouMind-OpenLab's awesome-gpt-image-2 repository has seen significant attention with a growth score of 5.02 and an impressive 3,423 stars, thanks to its curated library of GPT Image 2 prompts covering various languages and use cases. Researchers seeking to explore the capabilities of OpenAI's next-gen image model may find this resource invaluable.