DARASK Video Info

This is not a flashy node, and it doesn't need to be. When you load a video, everything downstream usually wants a plain number - an audio model wants the clip's duration in seconds, an interpolation node wants the source FPS, a resizer wants the loaded dimensions. DARASK Video Info takes the DARASK_VIDEO_INFO dict that DARASK Load Video (Upload) outputs and splits it into ten scalar outputs you can actually wire into anything. It's the VHS_VideoInfo equivalent for this pack, minus the confusion.

Source vs loaded

The ten outputs split into two obvious groups:

• source_fps, source_frame_count, source_duration, source_width, source_height - what the file actually is.
• loaded_fps, loaded_frame_count, loaded_duration, loaded_width, loaded_height - what you got after applying force_rate, select_every_nth, frame_load_cap and any custom resize on the loader.

That second group is the one that matters. If you resampled 30fps to 25fps or kept every 2nd frame, the source numbers lie about what the node actually loaded, and the loaded_* values are what your downstream models actually see.

What you'd reach for it for

The README's reference recipe: load a low-fps video → RIFE-interpolate it up to 25fps → generate audio with MMAudio that lines up with the motion. The critical wire is loaded_duration → MMAudioSampler's duration input - the audio length then matches the interpolated frame count exactly. You'd also grab loaded_fps to feed RIFE's source_fps so your interpolation math is based on the real post-filter rate, not the file's headline number.

Input: a single video_info socket. There's nothing to configure - it's a pure splitter, which is honestly the appeal. Node shows up, no sliders, no way to mis-set it.

Install

Same pack as everything else here: ComfyUI Manager → search DARASK → install, then restart. It needs the rest of the video half of the pack to be useful - by itself it's a dead end, since it has no video input of its own, only the loader's video_info output.

One note: it deliberately carries no audio pass-through. MMAudio workflows generate their own audio rather than reusing the source track, so the pack dropped audio from the loader side entirely. If you need the original audio preserved, this isn't the node chain for you - you'd want a VHS loader for that.