how to select a sd-model when use text2image api?

[shuaiyy]

I want to use a specific sd model when call the text2image api, how to to that?

I didn't find a field for that, likesd_model or ckpt_checksum or checkpoint ...

[Lebenoa]

Use OverrideSettings and DoNotOverrideSettingsRestoreAfterwards to set this value

Settings will restore to it's previous state (before sending the API) if DoNotOverrideSettingsRestoreAfterwards is not set or set to false

As you can see in the image the common one is sd_model_checkpoint. You can find this (the keys) in Web UI settings by toggle something on and off OR browsing thru config.json. The sd_model_checkpoint you can find in quick settings field.

If you want to I can make the field accept api.Options instead of map[string]any

P.S. Sorry for the informal and maybe rude answer. I'm not fluent in English

[Lebenoa]

I have update an example here or just view below

package main

import (
    "fmt"
    "os"

    "pg/webui-api"
    "pg/webui-api/sampler"
    "pg/webui-api/upscaler"
)

func main() {
    // Currently using `realdosmix_.safetensors [0d27c62ffa]` which is a realistic model
    API := api.New(api.Config{
        Default: &api.Default{
            Steps:    28,
            Sampler:  sampler.DPM_PLUS_PLUS_2M_KARRAS,
            Width:    384,
            Height:   512,
            CFGScale: 7.0,
        },
    })

    response, err := API.Text2Image(&api.Txt2Image{
        Prompt: api.BuildPrompt(
            "(masterpiece), (high quality), highres, (high resolution illustration), bishoujo, (game cg), intricate details, detailed_face",
            "1girl, skinny, narrow waist, white pantyhose",
            "(white serafuku), (school uniform), (blue sailor collar), (red neckerchief), (pleated skirt)",
            "large breasts, lipgloss, hair bow, (absurdly_long_hair:1.5)",
            "beautiful sky and beautiful garden, cowboy_shot",
        ),
        NegativePrompt:    "(KHFB, AuroraNegative), EasyNegative, badhandv4, (worst Quality, low quality:1.4), (multiple_girls:1.2), border, nsfw, skimpy, grayscale, umbrella",
        EnableHR:          true,
        HRScale:           2,
        HRUpscaler:        upscaler.R_ESRGAN_4x_PLUS_ANIME_6B,
        DenoisingStrength: 0.5,
        OverrideSettings: map[string]any{
            "sd_model_checkpoint": "AbyssOrangeMix2_hard", // a anime, semi-realistic style model
        },
        DecodeAfterResult: true,
    })
    pif(err)

    fmt.Println("Successfully decoded", len(response.DecodedImages), "images")
    for idx, decoded := range response.DecodedImages {
        file, err := os.Create(fmt.Sprint("override result", idx, ".png"))
        pif(err)

        file.Write(decoded)
        file.Close()
    }
}

func pif(err error) {
    if err != nil {
        panic(err)
    }
}

[shuaiyy]

thanks!

I'm try to build a group chatbot for AI drawing, and your SDK project can help me save a lot of time and coding.

If you want to I can make the field accept api.Options instead of map[string]any

It's a good idea to add a field, cause switching between different models while drawing is a very common scenario.

P.S. I am not a native English speaker, and some of my expressions were translated using ChatGPT.