Random Seed Batch Generator

Generate multiple groups of seeds for batch image processing. Each group produces similar variations by incrementing seeds with a fixed step. Use multiple groups to run several independent variation sets in a single execution.

Usage: Connect the seeds output to the seed input of any sampler node (e.g., KSampler, KSampler Advanced).

Inputs

| Parameter | Type | Default | Description | |-----------|------|---------|-------------| | initial_seed | INT | 0 | Starting seed for reproducible randomization. When set, the same initial_seed + control_after_generate will produce the same sequence of seeds. | | control_after_generate | - | randomize | Controls how initial_seed changes after each execution: fixed, increment, decrement, or randomize. | | step | INT | 1 | Step between seeds within a single group. Set to 0 for all seeds in a group to be equal. | | count | INT | 1 | Number of seeds per group. | | groups | INT | 1 | Number of groups to generate. Each group starts from a random seed. |

Outputs

| Output | Type | Description | |--------|------|-------------| | seeds | INT (list) | Flat list of all generated seeds. Total count = count * groups. |

How It Works

1. A deterministic RNG is seeded with initial_seed using random.Random(initial_seed).
2. For each group, a random starting seed is generated from this RNG in range 0..max_start.
3. Seeds within a group are calculated as: start + step + i for i in 0..count-1.
4. The starting seed is constrained so that all seeds in the group stay within the valid range 0..2^64-1.
5. All groups are concatenated into a single flat list.
6. After execution, initial_seed is updated according to control_after_generate: fixed (no change), increment (+1), decrement (-1), or randomize (new random value).

Example

With step=3, count=2, groups=5, the output will be 10 generations in batch:

• 5 pairs of similar generation variations
• In each pair the seed will differ by +3

With step=3, count=2, groups=5, the output might look like:

[12345, 12348, 67890, 67893, 11111, 11114, 22222, 22225, 33333, 33336]

Where:

• Group 1: [12345, 12348] (start=12345, step=3)
• Group 2: [67890, 67893] (start=67890, step=3)
• Group 3: [11111, 11114] (start=11111, step=3)
• Group 4: [22222, 22225] (start=22222, step=3)
• Group 5: [33333, 33336] (start=33333, step=3)

Note: Values are random and will differ on each execution.

Use Cases

• Generate similar image variations by using seeds with a small step.
• Run multiple independent variation sets in a single ComfyUI execution.
• Control how many generations to create within one workflow run.

Important Notes

• Reproducible with initial_seed: When initial_seed is set and control_after_generate is fixed or increment/decrement, the same sequence of seeds will be produced across executions.
• All seeds are guaranteed to be in range 0..2^64-1.
• If step + count - 1 > 2^64-1, the output will be empty (no valid seeds can be generated).
• Total output size = count * groups. Be careful with very large values (e.g., count=1000, groups=1000 produces 1,000,000 seeds).
• Batch size: The number of generated seeds (count * groups) determines the batch size when connected to KSampler. Make sure your KSampler batch_size matches or is compatible with this count.
• Empty output: If parameters are invalid (e.g., step + count - 1 > 2^64-1), the node returns an empty list, which may cause downstream errors.

License

This project is licensed under the GNU General Public License v3.0 (GPL-3.0). See LICENSE.txt for the full text.

Based on Seed Shifter from comfyui_steudio by Steudio, also licensed under GPL-3.0.

Links

• Repository: https://github.com/Glashkoff/comfyui-random-seed-batch-generator