Sammy Kumar
39d7164b0b
fix(media-selection): update error messages to reflect new media source options for clarity fix(llm_studio_structured.py): increase max_tokens default value to accommodate larger inputs test(redgifs): add unit tests for downloading media from Redgifs to ensure functionality and error handling test: add tests for direct download functionality from Redgifs to ensure reliability and correctness of media selection docs: update MediaSelection documentation to include support for direct Redgifs URLs in media sourcing options
1.3 KiB
1.3 KiB
MediaSelection
Select images or videos from uploads, filesystem directories, or Reddit sources, optionally trimming/resizing them and returning the resolved path plus metadata for downstream nodes.
Inputs
media_source– choose Upload, Randomize from Path, Reddit Post, or Randomize from Subreddit.media_type– image or video (controls downstream handling).seed– forces deterministic randomization when sampling folders/subreddits.- Optional fields:
media_path, upload fields, Reddit/subreddit URLs, max duration, resize mode + dimensions.
Outputs
media_path– absolute path to the selected media.media_type– echoes the resolved type (image/video).media_info– JSON summary string (dimensions, duration, FPS, resize decisions).height,width,duration,fps– numeric metadata for chaining into VHS/analysis nodes.
Usage Tips
- Use “Randomize Media from Path” with different seeds to iterate through a dataset quickly.
- “Auto (by orientation)” resize keeps common portrait/landscape sizes (832×480 or 480×832) without manual math.
- When sourcing Reddit media, provide either a full URL, a direct
Redgifs/GfycatURL, orr/subreddit—the node handles downloads and temp storage.