video_upscaler_per_chunk

Chunk-Based Video Upscaler

This workflow is designed to upscale long videos safely by processing only a selected chunk of frames at a time.

Instead of loading and upscaling an entire video at once, the workflow uses a chunk index system. You can specify:

• Chunk_Index: which part of the video to process

• Frame_Per_Chunk: how many frames to process per run

The workflow calculates the target frame range automatically, loads the selected portion of the video, upscales it with the selected upscale model, converts the processed frames back into a video, and saves the result as an MP4.

This is useful when a full-length video is too heavy to upscale in one pass, especially when using a high upscale factor such as 4x or 6x.

How to use:

1. Upload your input video.

2. Set Frame_Per_Chunk.

   • For example, 300 frames = 10 seconds at 30fps.

   • 150 frames = 5 seconds at 30fps.

3. Set Chunk_Index.

   • 0 = first chunk

   • 1 = second chunk

   • 2 = third chunk

   • and so on.

4. Run the workflow.

5. Repeat with the next Chunk_Index if you want to process the next part of the video.

The output is saved as a silent upscaled video chunk. Audio is intentionally left disconnected to avoid sync issues during chunk processing. If needed, the original audio can be added back after all chunks are processed and combined.

Recommended use:

• Use smaller Frame_Per_Chunk values for very high upscale factors or limited GPU memory.

• Use larger values if your environment can handle more frames at once.

• Rename the output file or filename prefix for each chunk to avoid overwriting previous results.

This workflow is intended as a reusable long-video upscaling template for cases where full-video upscaling is too memory-intensive.