[ComfyUI] Allow users to save/load test sessions for comparing results later without re-generating

🔗 Link to GitHub: [https://github.com/JasonHoku/ComfyUI-Ultimate-Auto-Sampler-Config-Grid-Testing-Suite](https://github.com/JasonHoku/ComfyUI-Ultimate-Auto-Sampler-Config-Grid-Testing-Suite) Or find it in Comfy Manager: ComfyUI-Ultimate-Auto-Sampler-Config-Grid-Testing-Suite Use the Builder Node to whip up your own iterations and easily test tons of models, loras, prompts, everything! Or just write or plug in some JSON and get a grid of results! It auto-generates grids based on your inputs (e.g., 3 samplers × 2 schedulers × 2 CFG x ALL LoRAs in FolderA in either each for each or combined!) and renders them in a zoomable, infinite-canvas dashboard. **The cool stuff:** * **Visual Config Builder**: A GUI to build your grids. Searchable dropdowns for models/LoRAs, drag sliders for strength, and easy toggles! * **Powerful Iteration Inputting:** Use arrays in JSON to run "each for each" iterations to display vast combinations of outputs rapidly with ease! Using a "\*" works for all samplers or all schedulers! * **Revise & Generate:** Click any image in the grid to tweak its specific settings and re-run just that one instantly. * **Session Saving:** Save/Load test sessions to compare results later without re-generating. * **Smart Caching:** Skips model re-loads so parameter tweaks are nearly instant. * **Curation:** Mark "bad" images with an X, and it auto-generates a clean JSON of only your accepted configs to copy-paste back into your workflow. * **Lightning Fast:** Splits up and batches tasks to minimize unloading and reloading models! * **Auto LoRA Triggers:** Automatically fetches trigger words from CivitAI (via hash lookup) and appends them to your prompts. You can even filter out specific triggers you don't want. * **Massive Scale:** Supports folder expansion (test ALL models or LoRAs in any folder), multi-LoRA stacking, and handles grids of thousands of images with virtual scrolling. * **Non-Standard Support: Works out of the box with SD3, Flux, Z-Image, etc. * **Smart Caching:** Skips reloading models/LoRAs if they are shared between consecutive runs. * **Resumable:** Stop a run halfway? It detects existing images and resumes where you left off. * **JSON Export:** Automatically formats your "Accepted" and "Favorite" images into clean JSON to copy-paste back into your workflow. A ton more features listed and explained on the readme on GitHub! **Repo:** [https://github.com/JasonHoku/ComfyUI-Ultimate-Auto-Sampler-Config-Grid-Testing-Suite](https://github.com/JasonHoku/ComfyUI-Ultimate-Auto-Sampler-Config-Grid-Testing-Suite) ## Here's some json_config examples you could plug in to instantly generate a variety of tests! # **Examples:** *This example generates 8 images (2 samplers × 2 schedulers × 2 steps × 1 cfg).* [ { "sampler": ["euler", "dpmpp_2m"], "scheduler": ["normal", "karras"], "steps": [20, 30], "cfg": [7.0, 8.0], "lora": "None", "str_model": 1.0, "str_clip": 1.0 } ] ## 🏆 Group 1: The "Gold Standards" (Reliable Realism) *Tests the 5 most reliable industry-standard combinations.* 5 samplers x 2 schedulers x 2 step settings x 2 cfgs = 40 images [ { "sampler": ["dpmpp_2m", "dpmpp_2m_sde", "euler", "uni_pc", "heun"], "scheduler": ["karras", "normal"], "steps": [25, 30], "cfg": [6.0, 7.0], "lora": "None", "str_model": 1.0, "str_clip": 1.0 } ] ## 🎨 Group 2: Artistic & Painterly *Tests 5 creative/soft combinations best for illustration and anime.* 5 samplers x 2 schedulers x 3 step settings x 3 cfgs = 90 images [ { "sampler": ["euler_ancestral", "dpmpp_sde", "dpmpp_2s_ancestral", "restart", "lms"], "scheduler": ["normal", "karras"], "steps": [20, 30, 40], "cfg": [5.0, 6.0, 7.0], "lora": "None", "str_model": 1.0, "str_clip": 1.0 } ] ## ⚡ Group 3: Speed / Turbo / LCM *Tests 4 ultra-fast configs. (Note: Ensure you are using a Turbo/LCM capable model or LoRA).* 4 samplers x 3 schedulers x 4 step settings x 2 cfgs = 96 images [ { "sampler": ["lcm", "euler", "dpmpp_sde", "euler_ancestral"], "scheduler": ["simple", "sgm_uniform", "karras"], "steps": [4, 5, 6, 8], "cfg": [1.0, 1.5], "lora": "None", "str_model": 1.0, "str_clip": 1.0 } ] ## 🦾 Group 4: Flux & SD3 Specials *Tests 4 configs specifically tuned for newer Rectified Flow models like Flux and SD3.* 2 samplers x 3 schedulers x 3 step settings x 2 cfgs = 36 images [ { "sampler": ["euler", "dpmpp_2m"], "scheduler": ["simple", "beta", "normal"], "steps": [20, 25, 30], "cfg": [1.0, 4.5], "lora": "None", "str_model": 1.0, "str_clip": 1.0 } ] ## 🧪 Group 5: Experimental & Unique *Tests 6 weird/niche combinations for discovering unique textures.* 6 samplers x 4 schedulers x 5 step settings x 4 cfgs = 480 images [ { "sampler": ["dpmpp_3m_sde", "ddim", "ipndm", "heunpp2", "dpm_2_ancestral", "euler"], "scheduler": ["exponential", "normal", "karras", "beta"], "steps": [25, 30, 35, 40, 50], "cfg": [4.5, 6.0, 7.0, 8.0], "lora": "None", "str_model": 1.0, "str_clip": 1.0 } ] I'd love to hear your feedback on it and if there's any other features that could be beneficial here!