Skip to content

AbstractVision v0.3.9

Choose a tag to compare

View all tags
@github-actions github-actions released this 21 May 21:57
· 25 commits to main since this release
v0.3.9
88e419a
  • Local Diffusers video: implement the first in-process text_to_video path through zai-org/CogVideoX-2b / THUDM/CogVideoX-2b, including registry-backed request normalization, preload warmup, unload support, MPS-safe explicit FP16 device moves, and MP4 artifact outputs packaged through ffmpeg.
  • Interactive surfaces: add one-shot abstractvision t2v, REPL /t2v, and playground Text → Video generation with backend/task-aware model discovery, progress reporting, and catalog surfacing for local video-capable models.
  • AbstractCore integration: route plugin t2v / i2v calls through the same explicit backend binding path used by image generation, extend residency task aliases for video, and track request-warm local text_to_video models in the loaded-model inventory.
  • Catalog/tests/docs: add curated capability/download metadata for zai-org/CogVideoX-2b, keep local image-edit surfaces intact, expand backend/CLI/playground/plugin regression coverage, refresh README/reference/FAQ/getting-started docs, and regenerate llms-full.txt.
Assets 4
Loading