← Back to all generators

ai-forever/kandinsky-2

text2img model trained on LAION HighRes and fine-tuned on internal datasets

Capabilities

1:1 4:3 3:4 16:9 9:16 2:3 3:2 Seed

Cost

Community model (estimated from hardware time)

Input Parameters

batch_size integer

Choose batch size. Lower the setting if out of memory.

Default: 1
1 2 3 4
guidance_scale number

Scale for classifier-free guidance

Default: 4 min: 1, max: 20
height integer

Choose height. Lower the setting if out of memory.

Default: 512
256 288 432 512 576 768 1024
num_inference_steps integer

Number of denoising steps

Default: 50 min: 1, max: 500
output_format string

Format of the output images

Default: "webp"
webp jpg png
output_quality integer

Quality of the output images, from 0 to 100. 100 is best quality, 0 is lowest quality.

Default: 80 min: 0, max: 100
prior_cf_scale integer
Default: 4
prior_steps string
Default: "5"
prompt string

Input Prompt

Default: "red cat, 4k photo"
scheduler string

Choose a scheduler

Default: "p_sampler"
ddim_sampler p_sampler plms_sampler
seed integer

Random seed. Leave blank to randomize the seed

width integer

Choose width. Lower the setting if out of memory.

Default: 512
256 288 432 512 576 768 1024
Version: 3c6374e7a9a1 Updated: 6/8/2026 6.2M runs