API support for Multi-ControlNet

[YanboFan]

How to use multi controlnet in the api mode? For example, I want to use both the control_v11f1p_sd15_depth and control_v11f1e_sd15_tile models. Has anyone tried this?

[YanboFan]

The following code works, https://github.com/Mikubill/sd-webui-controlnet/wiki/API#route-post-controlnetdetect

# A1111 payload
payload = {
    "prompt": 'your-promot',
    "negative_prompt": "",
    "batch_size": 1,
    "sampler_index": "Euler a",
    "seed":0,
    "steps": 20,
    "cfg_scale": 7,
    "alwayson_scripts": {
        "controlnet": {
            "args": [
                {
                    "input_image": LR_encoded_image,
                    "model": "control_v11f1e_sd15_tile [a371b31b]",
                    "resize_mode":1,
                    "control_mode":0,
                    "weight":1,
                },
                {
                    "input_image": Depth_encoded_image,
                    "model": "control_v11f1p_sd15_depth [cfd03158]",
                    "resize_mode":1,
                    "control_mode":0,
                    "weight":1,
                }
            ]
        }
    }
}

[Nomination-NRB]

The following code works, https://github.com/Mikubill/sd-webui-controlnet/wiki/API#route-post-controlnetdetect

# A1111 payload
payload = {
    "prompt": 'your-promot',
    "negative_prompt": "",
    "batch_size": 1,
    "sampler_index": "Euler a",
    "seed":0,
    "steps": 20,
    "cfg_scale": 7,
    "alwayson_scripts": {
        "controlnet": {
            "args": [
                {
                    "input_image": LR_encoded_image,
                    "model": "control_v11f1e_sd15_tile [a371b31b]",
                    "resize_mode":1,
                    "control_mode":0,
                    "weight":1,
                },
                {
                    "input_image": Depth_encoded_image,
                    "model": "control_v11f1p_sd15_depth [cfd03158]",
                    "resize_mode":1,
                    "control_mode":0,
                    "weight":1,
                }
            ]
        }
    }
}

when I use this payload, it will return 3 images, the same in the webui, but it seems that the softedge doesn't work.

using api:

using webui:

If I use only one controlnet, either openpose or softedge will work, but not both together

[aayushpanta]

I am facing the same issue. Using two contolnets with the api gives different results than the webui. Did you find a solution?

[Nomination-NRB]

I later found that one of my parameters was wrong. The default parameter on my ui: Mask blur is 4, but the parameter in my api is set to 0, and then the corresponding change to 4 is consistent with the ui result (if the graph is inconsistent, the probability is that there is no one-to-one correspondence between the parameters).

[aayushpanta]

Apparently, I had the same problem. Thanks!

[DevelopmentZheng]

@Nomination-NRB 你可以成功实现了两个带 api 的 contolnet 的结果吗

[Nomination-NRB]

@Nomination-NRB 你可以成功实现了两个带 api 的 contolnet 的结果吗

是的，之前失败是因为参数配错了，导致controlnet没有被使用

[DevelopmentZheng]

能把你的参数配置给一下吗

[Nomination-NRB]

详情见这个issue #13664