❌ Set Image [Deprecated]
The old image-geometry node that the empty-latent node ate
- genparams
- genparams
If you open an old TinyBreaker workflow, you'll meet ❌ Set Image [Deprecated] - the node that used to own everything about the initial canvas: seed, aspect ratio, orientation, size, and batch count, all packed into genparams. It's deprecated because the pack reorganized that whole job. The seed moved to its own node (Set Base Seed), and the geometry moved downstream into the latent generation, where Empty Latent Image reads aspect/size/batch straight out of the genparams blob. The old one-node-does-all approach didn't survive contact with the pack's "one setting, one node" redesign.
What it did
It was a one-stop image-setup node: seed (the initial noise pattern), ratio (a dropdown of aspect ratios from 1:1 square to 32:9 ultrawide - yes, including the gloriously absurd 70:27 cinerama), orientation (landscape/portrait), size (small/medium/large, where the tooltip notes medium is the training size but large is recommended), and batch_size (up to 4096). Output was the updated genparams. The seed it set is what Set Base Seed still sets today (denoising.base.noise_seed); the geometry keys are what Empty Latent Image now consumes.
Why it's gone
Splitting it made each concern independent and chainable. Want the same composition across several aspect ratios? In the old design you'd duplicate the whole node. Now Empty Latent Image pulls the ratio from the blob, so you can change aspect in one place (the unified prompt's --aspect flag, or whatever sets the geometry) and every downstream node follows. The size vocabulary also got normalized - small/medium/large gave way to explicit resolution handling, and the pack's unified-prompt parameters (--aspect, --landscape, --portrait, --medium) are the modern way to drive these same keys.
If you're stuck with one
The good news: it's mostly a rename. Replace Set Image with Set Base Seed (for the seed) and let Empty Latent Image handle geometry from the genparams chain; port batch_size to whatever sets image.batch_size in your pipeline (the unified prompt's --batch-size, or the genparams from the loader). If you relied on ratio's dropdown, the modern equivalents are the prompt parameters or the geometry keys - the aspect ratios themselves are identical. Old workflows keep loading thanks to the deprecated shim, but you're running a version of the pack's design that's been superseded, and porting forward is a few minutes of rewiring, not a redesign.
Install
Comes with the pack: ComfyUI Manager search "tinybreaker", or git clone https://github.com/martin-rizzo/ComfyUI-TinyBreaker into custom_nodes, restart. Deprecated nodes are hidden from the default menu, so you'll only encounter this one in saved graphs - and that's exactly the right way to meet it.
Inputs (6)
| Name | Type | Default | Description |
|---|---|---|---|
| genparams | GENPARAMS | The generation parameters to be updated. | |
| seed | INT | 11–18446744073709550000 | The pattern of the random noise to use as starting point for the image generation. (different seed numbers will generate totally different images) |
| ratio | COMBO | 1:1 (square) | The aspect ratio of the image. |
| orientation | COMBO | landscape | The orientation of the image. (landscape or portrait) |
| size | COMBO | large | The relative size for the image. ("medium" is the size the model was trained on, but "large" is recommended) |
| batch_size | INT | 11–4096 | The number of images to generate in a single batch. |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| genparams | GENPARAMS | The generation parameters updated with the new image attributes. (you can use this output to chain other genparams nodes) |