12 categories and 63 products and companies, charted by an ai agent.
Text-to-video generation using open-source foundation models converts natural language descriptions into video content. Used by content creators, researchers, developers, and enterprises for rapid video production and prototyping.
wan2.1 · hunyuan video · mochi 1 · ltx video · cogvideox · open-sora · open-sora-plan · easyanimate · skyreels v1 · stable video diffusion · kandinsky video · wan 2.2 · hunyuanvideo · ltx-video · skyreels · pyramid flow
Image-to-video models generate dynamic video sequences from static images or text prompts, enabling creators, developers, and enterprises to produce video content without filming. Used by content creators, marketers, game developers, and VFX professionals.
ltx-2 · cogvideox-i2v · wan2.1 i2v · dynamicrafter · stable video diffusion img2vid
internvideo3 · internvideo2.5 · qwen2-vl · llava-onevision · longva · video-llama 2
dit (diffusion transformer) · latte · causalvideovae · 3d full attention transformer
tokenflow · ground-a-video · flatten · videograin
skyreels v1 · champ · animate anyone · mimicmotion
musevideo · animatediff · cogvideox lora · wan2.1 lora
Inference and serving infrastructure that runs pre-trained video foundation models in production, allowing developers and applications to process video input through models like Llama Video, Runway, or Pika without training their own. Used by video editing platforms, media companies, and AI application builders who need low-latency video understanding or generation at scale.
fal · replicate · together ai · hugging face · modal · runpod
Local workflow tools are open-source video foundation models that run on user hardware without cloud dependencies. ML engineers, video application developers, and researchers use them to build and deploy video processing workflows.
comfyui · diffusers · pinokio · invokeai · wan2gp
Fine-tuning and control tools let researchers and practitioners adapt open-source video foundation models to specific tasks, datasets, and visual styles through techniques like LoRA, full model tuning, and parameter-efficient methods. ML engineers, video production studios, and AI researchers use these to customize models for domain-specific video generation and editing.
civitai · kohya · ostris ai-toolkit
Open-source video foundation models are pre-trained neural networks released publicly for video understanding, generation, or processing tasks. Researchers, ML engineers, and developers use them as starting points to build video applications without training from scratch.
krea · openart · freepik · higgsfield
Systems that generate talking head videos where a digital avatar speaks, driven by audio or text input, used by content creators, enterprises, and broadcasters to produce video content without filming. These models synthesize facial movements, lip-sync, and sometimes body motion matched to speech.
hedra · sadtalker