DARASK Exif Read

The Exif Apply nodes are the heavy hitters - they read metadata and load models and encode prompts. But sometimes you just want the facts without any of that happening. DARASK Exif Read is the stripped-down parser: same metadata engine, zero model loading, zero VRAM. Point it at a filepath and it hands you every recoverable field as a plain output you can route anywhere. It's the node you use when you want to inspect what an image was made with, log it, or wire individual values into a workflow that builds its own pipeline.

Why you'd use it over the Apply variants

The Apply variants are opinionated - they assume you want the full rebuilt pipeline. Exif Read assumes nothing, which makes it better for three jobs:

1. Debugging. Ask "what sampler and CFG did this image actually use?" without triggering a multi-GB model load just to find out.
2. Custom pipelines. Recover positive and negative as plain strings and feed them into your own prompt-handling chain - Prompt Cells, a different encoder, a pre-sampler you built yourself.
3. Logging / organization. Wire model_name and seed into a Show Text or a filename builder to keep records of what's in a folder.

It's also a nice sanity check before you commit to a batch run: drop it on a sample image, confirm the metadata you're expecting is actually there, then swap in the Apply node for real work.

Inputs and outputs

One required input, filepath. Outputs are all passive reads: positive and negative (strings), loras_text, model_name, vae_name, seed, cfg, sampler_name, scheduler, steps, denoise, width, height, and raw_metadata - the unparsed metadata blob if you want to eyeball or grep it yourself.

Note what's not here compared to the Apply nodes: no model, clip or vae outputs, no CONDITIONING, no loras_applied/positive_text distinction - loras_text is the recovered LoRA list as text. If you find yourself rebuilding all of that downstream by hand, you've probably outgrown this node and should just use an Apply variant. It's a read-only tool by design.

Install

The usual DARASK pack route: ComfyUI Manager → search DARASK → install, restart. piexif comes with the pack's requirements. Nothing to download - it parses whatever you already have.

One honest caveat: metadata quality is whatever the source put in the file. A1111 and ComfyUI outputs carry rich info, but stripped or resaved images will come back mostly empty. The node can't invent fields that aren't there, and raw_metadata will happily show you exactly how thin a file is.