Nodes/ComfyUI-OllamaOmni/Ollama Video Analyze
ComfyUI Node

Ollama Video Analyze

Sample frames from a video and analyze them with an Ollama vision model. Works with any IMAGE batch representing a frame sequence.

By ckinpdx·Created 4 months ago·Updated 4 months ago· 1
Ollama Video Analyze
  • frames
  • connectivity
  • options
  • meta
  • result
  • thinking
  • meta
promptDescribe what happens in this video clip.
max_frames8
sample_mode
systemYou are analyzing frames sampled from a video clip. Describe the action, subjects, and content.
thinkfalse
format
CategoryOllama/Video

Inputs (10)

NameTypeDefaultDescription
framesIMAGEVideo frames as an IMAGE batch (e.g. from VHS Load Video or any frame sequence node).
promptSTRINGDescribe what happens in this video clip.Prompt for the vision model.
max_framesINT81–64Maximum frames to sample and send. More = richer context but slower and higher memory.
sample_modeCOMBOSampling strategy: uniform=evenly spaced across entire clip, first=opening frames, last=closing frames.
systemoptSTRINGYou are analyzing frames sampled from a video clip. Describe the action, subjects, and content.System prompt for the model.
thinkoptBOOLEANfalseEnable chain-of-thought reasoning (model must support this).
formatoptCOMBOOutput format.
connectivityoptOLLAMA_CONNECTIVITY
optionsoptOLLAMA_OPTIONS
metaoptOLLAMA_META

Outputs (3)

NameTypeDescription
resultSTRING
thinkingSTRING
metaOLLAMA_META