Today's Image & Video Generation: Fastest-Growing Projects — April 21, 2026
This week, the Image & Video Generation space saw significant growth, driven by advancements in multimodal foundation models and text-to-image generation. Several repositories gained traction, showcasing innovative approaches to image understanding, editing, and video rendering. Notably, many of these tools focus on leveraging AI-native technologies to enhance visual content creation.
JoyAI-Image, with a Growth Score of 61.07 and 1,938 stars, is a unified multimodal foundation model for image understanding, text-to-image generation, and instruction-guided image editing. Its rapid growth can be attributed to its comprehensive capabilities, making it an attractive choice for developers seeking an all-in-one solution.
ERNIE-Image, developed by the ERNIE-Image team at Baidu, boasts a Growth Score of 36.64 and 321 stars. This open text-to-image generation model reaches state-of-the-art performance among open-weight models, thanks to its single-stream Diffusion Transformer (DiT) architecture with only 8B DiT parameters.
Ashim-hq's ashim repository has gained significant attention, with a Growth Score of 33.04 and 845 stars. This Docker container packs over 30 tools for image manipulation, including resizing, compression, background removal, and OCR, all without relying on cloud services or telemetry. Its growth is likely driven by the demand for self-contained, AI-powered image processing solutions.
Kangarooking's design-image-studio has a Growth Score of 30.83 and 77 stars. This tool converts vague visual demands into high-quality, actionable design images. Although its commit activity is relatively low, its unique approach to visual content creation has sparked interest among developers.
MOSS-VL, with a Growth Score of 17.73 and 227 stars, is the core multimodal model series within the OpenMOSS ecosystem. Its focus on visual understanding and commitment to open research have contributed to its growth, as developers seek to leverage its capabilities for various applications.
ShandaAI's AlayaRenderer boasts a Growth Score of 16.47 and 569 stars. This AI-native Renderer is designed specifically for games and virtual worlds, offering a unique solution for high-performance rendering. Its growth is likely driven by the increasing demand for immersive experiences in gaming and virtual reality.
F-R-L's forge-film has a Growth Score of 13.62 and 508 stars. This multi-model DAG-driven parallel AI film generation engine enables simultaneous scene generation, significantly speeding up production. Its innovative approach to film creation has garnered attention from developers seeking to streamline their workflows.
Gulucaptain's Camera-Transformer-1 has gained traction with a Growth Score of 11.64 and 106 stars. This tool allows users to generate videos with precise camera control, understanding intent and generating spatially-aware footage. Its growth is likely driven by the demand for more sophisticated video creation tools.
Jiangchaokang's awesome-generative-models repository has a Growth Score of 8.21 and 100 stars. Although not directly related to image or video generation, this collection of diffusion models provides valuable resources for developers working in the field.
Lastly, NVlabs' PixelDiT, with a Growth Score of 4.18 and 154 stars, focuses on pixel diffusion transformers for image generation. Despite its relatively low growth score, this repository has garnered interest due to its innovative approach to image creation and its presentation as an oral at CVPR 2026.
JoyAI-Image, with a Growth Score of 61.07 and 1,938 stars, is a unified multimodal foundation model for image understanding, text-to-image generation, and instruction-guided image editing. Its rapid growth can be attributed to its comprehensive capabilities, making it an attractive choice for developers seeking an all-in-one solution.
ERNIE-Image, developed by the ERNIE-Image team at Baidu, boasts a Growth Score of 36.64 and 321 stars. This open text-to-image generation model reaches state-of-the-art performance among open-weight models, thanks to its single-stream Diffusion Transformer (DiT) architecture with only 8B DiT parameters.
Ashim-hq's ashim repository has gained significant attention, with a Growth Score of 33.04 and 845 stars. This Docker container packs over 30 tools for image manipulation, including resizing, compression, background removal, and OCR, all without relying on cloud services or telemetry. Its growth is likely driven by the demand for self-contained, AI-powered image processing solutions.
Kangarooking's design-image-studio has a Growth Score of 30.83 and 77 stars. This tool converts vague visual demands into high-quality, actionable design images. Although its commit activity is relatively low, its unique approach to visual content creation has sparked interest among developers.
MOSS-VL, with a Growth Score of 17.73 and 227 stars, is the core multimodal model series within the OpenMOSS ecosystem. Its focus on visual understanding and commitment to open research have contributed to its growth, as developers seek to leverage its capabilities for various applications.
ShandaAI's AlayaRenderer boasts a Growth Score of 16.47 and 569 stars. This AI-native Renderer is designed specifically for games and virtual worlds, offering a unique solution for high-performance rendering. Its growth is likely driven by the increasing demand for immersive experiences in gaming and virtual reality.
F-R-L's forge-film has a Growth Score of 13.62 and 508 stars. This multi-model DAG-driven parallel AI film generation engine enables simultaneous scene generation, significantly speeding up production. Its innovative approach to film creation has garnered attention from developers seeking to streamline their workflows.
Gulucaptain's Camera-Transformer-1 has gained traction with a Growth Score of 11.64 and 106 stars. This tool allows users to generate videos with precise camera control, understanding intent and generating spatially-aware footage. Its growth is likely driven by the demand for more sophisticated video creation tools.
Jiangchaokang's awesome-generative-models repository has a Growth Score of 8.21 and 100 stars. Although not directly related to image or video generation, this collection of diffusion models provides valuable resources for developers working in the field.
Lastly, NVlabs' PixelDiT, with a Growth Score of 4.18 and 154 stars, focuses on pixel diffusion transformers for image generation. Despite its relatively low growth score, this repository has garnered interest due to its innovative approach to image creation and its presentation as an oral at CVPR 2026.