Consistent Text-to-Image Generation on ConsiStory+ (test)
0.9074CLIP-TVanilla SDXL
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Vanilla SDXLSteps=50, Memory (GB)=16.04, Inference Time (s)=13.0892026.02 | 0.9074 | 0.8165 | 0.5292 | — | — | — | — | — | |
| DeCorStoryBase Model=SDXL, Train-Free=true, Steps=50, Memory (GB)=18.72, Inference Time (s)=23.35212026.02 | 0.9001 | 0.9134 | 0.1922 | — | — | — | — | — | |
| SDeC+ConsiStoryBase model=SDXL, PE=true, Steps=502025.10 | 0.8967 | 0.8744 | — | 0.2542 | 0.4155 | 0.67 | 27 | 15.56 | |
| SDeCBase model=SDXL, PE=true, Steps=502025.10 | 0.8946 | 0.8655 | — | 0.2589 | 0.3675 | 0.61 | 15 | 12.14 | |
| 1Prompt1StoryBase Model=SDXL, Train-Free=true, Steps=50, Memory (GB)=18.7, Inference Time (s)=23.20882026.02 | 0.8942 | 0.9117 | 0.1993 | — | — | — | — | — | |
| ConsiStoryBase model=SDXL, PE=false, Steps=502025.10 | 0.8942 | 0.8604 | — | 0.2729 | 0.4207 | 0 | 27 | 15.58 | |
| 1P1SBase model=SDXL, PE=false, Steps=502025.10 | 0.8883 | 0.8798 | — | 0.2238 | 0.2955 | 0.1 | 22 | 13.1 | |
| StoryDiffusionBase Model=SDXL, Train-Free=true, Steps=50, Memory (GB)=45.61, Inference Time (s)=25.69282026.02 | 0.8877 | 0.8755 | 0.3212 | — | — | — | — | — | |
| Baseline (SDXL)Base model=SDXL, PE=false, Steps=502025.10 | 0.8865 | 0.8558 | — | 0.2778 | 0.3861 | 0 | 9 | 10.72 | |
| PhotoMakerBase model=SDXL, PE=false, Steps=502025.10 | 0.8812 | 0.8545 | — | 0.2808 | 0.3957 | 0 | 9 | 9.74 | |
| ConsiStoryBase Model=SDXL, Train-Free=true, Steps=50, Memory (GB)=34.55, Inference Time (s)=34.58942026.02 | 0.8769 | 0.8737 | 0.3188 | — | — | — | — | — | |
| PhotoMakerBase Model=SDXL, Train-Free=false, Steps=50, Memory (GB)=23.79, Inference Time (s)=18.02592026.02 | 0.8651 | 0.8465 | 0.3996 | — | — | — | — | — | |
| 1P1S w/o IPCABase model=SDXL, PE=false, Steps=502025.10 | 0.8637 | 0.8617 | — | 0.2682 | 0.3338 | 0.1 | 19 | 10.74 | |
| StoryDiffusionBase model=SDXL, PE=false, Steps=502025.10 | 0.8578 | 0.8502 | — | 0.3197 | 0.4214 | 0 | 24 | 38.44 | |
| Textual InversionBase model=SDXL, PE=false, Steps=402025.10 | 0.8557 | 0.8437 | — | 0.3066 | 0.3919 | 0 | 10 | 14 | |
| IP-AdapterBase Model=SDXL, Train-Free=false, Steps=30, Memory (GB)=19.39, Inference Time (s)=13.45942026.02 | 0.8458 | 0.9429 | 0.1462 | — | — | — | — | — | |
| Naive Prompt Reweighting (NPR)Base Model=SDXL, Train-Free=true, Steps=50, Memory (GB)=16.04, Inference Time (s)=17.24132026.02 | 0.8411 | 0.8916 | 0.2548 | — | — | — | — | — | |
| Textual InversionBase Model=SD1.5, Train-Free=false, Steps=40, Memory (GB)=32.94, Inference Time (s)=282.5072026.02 | 0.8378 | 0.8229 | 0.4268 | — | — | — | — | — | |
| Vanilla SD1.5Steps=50, Memory (GB)=4.73, Inference Time (s)=2.46572026.02 | 0.8353 | 0.7474 | 0.5873 | — | — | — | — | — | |
| Baseline (SD1.5)Base model=SD1.5, PE=false, Steps=502025.10 | 0.8324 | 0.8071 | — | 0.4118 | 0.4673 | 0 | 2 | 3.18 | |
| BLIP-DiffusionBase model=SD1.5, PE=false, Steps=262025.10 | 0.8187 | 0.8522 | — | 0.2851 | 0.3957 | 0 | 1 | 3.54 | |
| The Chosen OneBase Model=SDXL, Train-Free=false, Steps=35, Memory (GB)=10.93, Inference Time (s)=11.20732026.02 | 0.7614 | 0.7831 | 0.4929 | — | — | — | — | — | |
| BLIP-DiffusionBase Model=SD1.5, Train-Free=false, Steps=26, Memory (GB)=7.75, Inference Time (s)=1.92842026.02 | 0.7607 | 0.8863 | 0.283 | — | — | — | — | — |