cinemasetfree.com
cjwbw/supir
Practicing Model Scaling for Photo-Realistic Image Restoration In the Wild. This version uses LLaVA-13b for captioning.
Capabilities
Cost
Community model (estimated from hardware time)
Input Parameters
| Name | Type | Description | Default | Constraints |
|---|---|---|---|---|
image * | string (uri) | Low quality input image. | — | — |
a_prompt | string | Additive positive prompt for the inputs. | "Cinematic, High Contrast, highly detailed, taken using a Canon EOS R camera, hyper detailed photo - realistic maximum detail, 32k, Color Grading, ultra HD, extreme meticulous detailing, skin pore detailing, hyper sharpness, perfect without deformations." | — |
color_fix_type | string | Color Fixing Type.. | "Wavelet" | None AdaIn Wavelet |
edm_steps | integer | Number of steps for EDM Sampling Schedule. | 50 | min: 1, max: 500 |
linear_CFG | boolean | Linearly (with sigma) increase CFG from 'spt_linear_CFG' to s_cfg. | false | — |
linear_s_stage2 | boolean | Linearly (with sigma) increase s_stage2 from 'spt_linear_s_stage2' to s_stage2. | false | — |
min_size | number | Minimum resolution of output images. | 1024 | — |
model_name | string | Choose a model. SUPIR-v0Q is the default training settings with paper. SUPIR-v0F is high generalization and high image quality in most cases. Training with light degradation settings. Stage1 encoder of SUPIR-v0F remains more details when facing light degradations. | "SUPIR-v0Q" | SUPIR-v0Q SUPIR-v0F |
n_prompt | string | Negative prompt for the inputs. | "painting, oil painting, illustration, drawing, art, sketch, oil painting, cartoon, CG Style, 3D render, unreal engine, blurring, dirty, messy, worst quality, low quality, frames, watermark, signature, jpeg artifacts, deformed, lowres, over-smooth" | — |
s_cfg | number | Classifier-free guidance scale for prompts. | 7.5 | min: 1, max: 20 |
s_churn | number | Original churn hy-param of EDM. | 5 | — |
s_noise | number | Original noise hy-param of EDM. | 1.003 | — |
s_stage1 | integer | Control Strength of Stage1 (negative means invalid). | -1 | — |
s_stage2 | number | Control Strength of Stage2. | 1 | — |
seed | integer | Random seed. Leave blank to randomize the seed | — | — |
spt_linear_CFG | number | Start point of linearly increasing CFG. | 1 | — |
spt_linear_s_stage2 | number | Start point of linearly increasing s_stage2. | 0 | — |
upscale | integer | Upsampling ratio of given inputs. | 1 | — |
use_llava | boolean | Use LLaVA model to get captions. | true | — |
image required string Low quality input image.
a_prompt string Additive positive prompt for the inputs.
"Cinematic, High Contrast, highly detailed, taken using a Canon EOS R camera, hyper detailed photo - realistic maximum detail, 32k, Color Grading, ultra HD, extreme meticulous detailing, skin pore detailing, hyper sharpness, perfect without deformations." color_fix_type string Color Fixing Type..
"Wavelet" edm_steps integer Number of steps for EDM Sampling Schedule.
50 min: 1, max: 500 linear_CFG boolean Linearly (with sigma) increase CFG from 'spt_linear_CFG' to s_cfg.
false linear_s_stage2 boolean Linearly (with sigma) increase s_stage2 from 'spt_linear_s_stage2' to s_stage2.
false min_size number Minimum resolution of output images.
1024 model_name string Choose a model. SUPIR-v0Q is the default training settings with paper. SUPIR-v0F is high generalization and high image quality in most cases. Training with light degradation settings. Stage1 encoder of SUPIR-v0F remains more details when facing light degradations.
"SUPIR-v0Q" n_prompt string Negative prompt for the inputs.
"painting, oil painting, illustration, drawing, art, sketch, oil painting, cartoon, CG Style, 3D render, unreal engine, blurring, dirty, messy, worst quality, low quality, frames, watermark, signature, jpeg artifacts, deformed, lowres, over-smooth" s_cfg number Classifier-free guidance scale for prompts.
7.5 min: 1, max: 20 s_churn number Original churn hy-param of EDM.
5 s_noise number Original noise hy-param of EDM.
1.003 s_stage1 integer Control Strength of Stage1 (negative means invalid).
-1 s_stage2 number Control Strength of Stage2.
1 seed integer Random seed. Leave blank to randomize the seed
spt_linear_CFG number Start point of linearly increasing CFG.
1 spt_linear_s_stage2 number Start point of linearly increasing s_stage2.
0 upscale integer Upsampling ratio of given inputs.
1 use_llava boolean Use LLaVA model to get captions.
true 1302b550b4f7 Updated: 2/26/2026 190.5K runs