TS Studio Input (Seed)

Seeds are the fiddliest parameter in any ComfyUI workflow: they need to reach the sampler, they need to change on every run if you're iterating, and if you forget to convert the seed widget to an input the whole "new seed each run" ritual quietly stops working. In a TS Image Studio backend, the seed is a first-class control in the app's UI - and TS Studio Input (Seed) is the marker that carries it into the graph. It's a placeholder the studio sets on every run (manual or randomized), and you wire its INT output straight into the sampler's seed.

How it works

The standard marker pattern, trimmed to the essentials:

• param_name (default seed) - the identifier the studio uses to find this marker and set its value. Must be unique per workflow.
• value - an INT (default 0, full 64-bit seed range) that the studio overwrites. When you press the studio's randomize button or type a seed, this is what changes.

There's no label input on this one - the seed is such a well-understood parameter that the default name is fine. The single output is seed (INT), which you connect to the sampler's seed input. In the current ComfyUI world, that means wiring it into the noise generator - the README's own advice is to connect it to RandomNoise/KSampler seed - so every sampler in the graph that should follow the studio's seed is driven from one authoritative source.

The key difference from a hand-typed seed: the studio sets this value on every run, manual or randomized. You never have to touch the graph between runs - the app handles the ritual for you. And like every studio marker, loaded by itself in plain ComfyUI it just outputs whatever value is in the field, so a backend workflow is still runnable by hand with a fixed seed.

Install and use

No extra dependencies - part of comfyui-timesaver.

cd ComfyUI/custom_nodes
git clone https://github.com/AlexYez/comfyui-timesaver
cd comfyui-timesaver
python -m pip install -r requirements.txt

As a backend builder: give it the unique param_name (seed by convention), wire the seed output to every sampler that should follow the studio's control, and let the app fill the value. As a user who opened a workflow and found this node: don't fight it - the studio owns that value.

Common issues

• "Same result every run." The studio set a fixed seed (that's a feature - reproducibility). Use its randomize control, or check whether this marker's value is actually connected to the sampler; a sampler with a baked-in widget seed ignores this node entirely.
• "Changing the seed in the graph does nothing." Correct - the studio overwrites it at run time. The studio's UI is the source of truth while the app runs.
• "Two samplers, one seed." That's the point: wire both samplers to this one marker and they stay in lockstep.

The honest scope: it's the smallest of the markers - one INT in, one INT out - but it's the one that makes the studio's "run a hundred variations without touching the graph" workflow actually work.