ES Guides 7
The unglamorous glue node that powers ES Translate
- src_img_0
- tgt_img_0
- src_img_1
- tgt_img_1
- src_img_2
- tgt_img_2
- src_img_3
- tgt_img_3
- src_img_4
- tgt_img_4
- src_img_5
- tgt_img_5
- src_img_6
- tgt_img_6
- src_imgs
- tgt_imgs
- wgts
ES Guides 7 is the boring node in the ComfyUI-EbSynth pack, and that's the point. It doesn't run any synthesis itself - no dll, no optical flow, nothing clever. It's a packer: it takes up to seven guide pairs and hands them to ES Translate in exactly the format that node wants. If you've only ever used ES Translate with its three basic images, this is the node that unlocks the full EbSynth guide mechanism without you having to think about how the data gets passed.
What it actually does
Classic EbSynth workflows aren't limited to "this image should look like that image." They use guides: extra pairs of (source, target) images that steer the patch matching - a depth map, a directional light map, an edge map, each with its own strength. The original app's signature stylit example stacks two guides at weight 0.66 each on top of the base style. ES Guides 7 is how you do that in ComfyUI.
You get seven slots (src_img_0 … src_img_6 and matching tgt_img_0 … tgt_img_6, each with a weight_0 … weight_6 float). Slot zero is required; the rest are optional and skipped when left empty. The node concatenates all the sources into one batch, all the targets into another, and serializes the weights into a comma-separated string. That's the whole job.
The three outputs
- src_imgs - the batched guide source images (IMAGE)
- tgt_imgs - the batched guide target images (IMAGE)
- wgts - the weights as a STRING, e.g.
"1.0,0.66,0.66"
Wire all three into ES Translate's optional src_imgs, tgt_imgs, and wgts inputs, and you're done. If you're not using ES Translate, this node has nothing to offer you - it's pure glue.
The one thing that trips people up
wgts is a string, not a number, and not a tensor. The serialization is invisible to you - ES Guides 7 makes it and ES Translate consumes it - so you mostly don't care, except when you try to bypass the node and hand-roll the input. Build the string yourself and it has to be comma-separated floats in the same order as the images. Easier to just keep the node in the workflow.
There's no tuning here, no gotchas, nothing to troubleshoot. Install it as part of the pack - ComfyUI Manager search "ComfyUI-EbSynth", or git clone --recurse-submodules https://github.com/FuouM/ComfyUI-EbSynth.git in custom_nodes plus pip install -r requirements.txt - and it just sits there being useful. It's the sort of node that makes a workflow readable: seven labeled guide slots instead of a rat's nest of hand-packed tensors, which is exactly what you want when you're debugging why a style transfer is drifting.
Inputs (21)
| Name | Type | Default | Description |
|---|---|---|---|
| src_img_0 | IMAGE | — | |
| tgt_img_0 | IMAGE | — | |
| weight_0 | FLOAT | 1.00 | — |
| src_img_1opt | IMAGE | — | |
| tgt_img_1opt | IMAGE | — | |
| weight_1opt | FLOAT | 1.00 | — |
| src_img_2opt | IMAGE | — | |
| tgt_img_2opt | IMAGE | — | |
| weight_2opt | FLOAT | 1.00 | — |
| src_img_3opt | IMAGE | — | |
| tgt_img_3opt | IMAGE | — | |
| weight_3opt | FLOAT | 1.00 | — |
| src_img_4opt | IMAGE | — | |
| tgt_img_4opt | IMAGE | — | |
| weight_4opt | FLOAT | 1.00 | — |
| src_img_5opt | IMAGE | — | |
| tgt_img_5opt | IMAGE | — | |
| weight_5opt | FLOAT | 1.00 | — |
| src_img_6opt | IMAGE | — | |
| tgt_img_6opt | IMAGE | — | |
| weight_6opt | FLOAT | 1.00 | — |
Outputs (3)
| Name | Type | Description |
|---|---|---|
| src_imgs | IMAGE | — |
| tgt_imgs | IMAGE | — |
| wgts | STRING | — |