ComfyUI Node
Inference Time Scaler
Performs inference-time optimization to find the best image matching your text prompt. Supports two search algorithms: 1. Random Search: Generates multiple images with different random seeds 2. Zero-Order Search: Performs gradient-free local optimization Uses an ensemble of AI verifiers (CLIP, ImageReward, Qwen-VL) to score and rank the generated images. Returns the best image, a grid of top-k results, and detailed scoring information.
Inference Time Scaler
- model
- vae
- positive
- negative
- latent_image
- loaded_clip_score_verifier
- loaded_image_reward_verifier
- loaded_qwen_verifier
- Best Image
- Top-k Grid
- Top-k Score(s)
◄seed0►
◄steps20►
◄cfg1.0►
◄sampler_name▾►
◄scheduler▾►
◄denoise1.00►
◄search_algorithmrandom►
◄text_prompt_to_compare►
◄search_rounds5►
◄view_top_k3►
◄num_neighbors4►
◄lambda_threshold0.90►
CategoryInferenceTimeScaling
Inputs (20)
| Name | Type | Default | Description |
|---|---|---|---|
| model | MODEL | Denoising model. | |
| vae | VAE | VAE model for decoding latents. | |
| seed | INT | 00–18446744073709550000 | Random seed. |
| steps | INT | 201–10000 | Number of denoising steps to apply during each forward evaluation. |
| cfg | FLOAT | 1.00–100 | Classifier-Free Guidance Scale. |
| sampler_name | COMBO | Sampling algorithm to be used during each forward evaluation. | |
| scheduler | COMBO | Noise removal scheduler. | |
| positive | CONDITIONING | Positive prompt conditioning. | |
| negative | CONDITIONING | Negative prompt conditioning. | |
| latent_image | LATENT | Latent image to be denoised. | |
| denoise | FLOAT | 1.000–1 | Amount of denoising. |
| search_algorithm | COMBO | random | Select the search algorithm: 'random' for standard random search or 'zero-order' for gradient-free local search. |
| text_prompt_to_compare | STRING | Text prompt for verifier(s). | |
| search_rounds | INT | 51–10000 | Number of search rounds (random seeds for random search, or iterations for zero-order search). |
| view_top_k | INT | 31–100 | Return grid view of the top-k images. |
| num_neighbors | INT | 41–100 | Number of neighbors to sample per iteration in zero-order search (only used if search_algorithm is 'zero-order'). |
| lambda_threshold | FLOAT | 0.900–1 | Perturbation step size for zero-order search (only used if search_algorithm is 'zero-order'). |
| loaded_clip_score_verifieropt | CS_VERIFIER | HF CLIP model identifier. | |
| loaded_image_reward_verifieropt | IR_VERIFIER | ImageReward model identifier. | |
| loaded_qwen_verifieropt | QWN_VERIFIER | Qwen model identifier. |
Outputs (3)
| Name | Type | Description |
|---|---|---|
| Best Image | IMAGE | Best single image, grid of the top-k images, and their scores in JSON. |
| Top-k Grid | IMAGE | — |
| Top-k Score(s) | STRING | — |