DD 5.6 Settings | Coherent characters | Attempt 2

Angel.md

Attempt to generate a coherent character Part 2

I'm still looking for the right settings, and might begin with my previous attempts with a kitsune bard.

Of the three examples below, most break the ["description", "artist", "modifiers"] list convention, and all are relatively uncluttered descriptions.

Primary Settings

Setting	TheHistorian	WAS
diffusion_model	portrait_generator_v001	custom
clip_guidance_scale	5000	2000
tv_scale	50000	100
range_scale	50000	150
sat_scale	50000	0
cutn_batches	1	1
clamp_grad	true	true
clamp_max	0.1	0.05
seed	622116653	3138
eta	0.9	0.8

So, what do we learn from this? Nothing yet.

I'll run three batches side-by-side and see which ones get the clearest result with a prompt altered for an Angel.

Example1.md

An example of coherence in a prompt by VincentVip

The following is a fantastic example of portrait coherence using a simple prompt by VincentVip in the DD Discord server's pwompt-land channel.

Full-size image here

The prompt

"text_prompts": {
        "0": [
            "An ultradetailed portrait of a feamle northern europe goddess of ice, by tian zi trending on artstation :1.7",
            "cinematic lightsthomas kinkade, 4k octane render, artstation, pinterest"
        ]
    }

Example2.md

Example by TheHistorian

This is a much more realistic and simple portrait by TheHistorian on the DD Discord's pwompt_land channel.

Note that features outside the hair and face are not rendered by the model, so this may be less relevant for characters with wings.

Full-size image here

The full settings

{
    "text_prompts": {
        "0": [
            "a portrait photograph of a female by Alessio Albi:3",
            "beautiful full face, symmetrical face:1.5",
            "artstation,deviantart,hyperrealism:1.5",
            "green eyes, long red hair,a small nose:1"
        ]
    },
    "image_prompts": {},
    "clip_guidance_scale": 5000,
    "tv_scale": 50000,
    "range_scale": 50000,
    "sat_scale": 50000,
    "cutn_batches": 1,
    "max_frames": 10000,
    "interp_spline": "Linear",
    "init_image": null,
    "init_scale": 1000,
    "skip_steps": 0,
    "frames_scale": 1500,
    "frames_skip_steps": "60%",
    "perlin_init": false,
    "perlin_mode": "mixed",
    "skip_augs": false,
    "randomize_class": true,
    "clip_denoised": false,
    "clamp_grad": true,
    "clamp_max": 0.1,
    "seed": 622116653,
    "fuzzy_prompt": false,
    "rand_mag": 0.05,
    "eta": 0.9,
    "width": 512,
    "height": 640,
    "diffusion_model": "portrait_generator_v001",
    "use_secondary_model": false,
    "steps": 250,
    "diffusion_steps": 1000,
    "diffusion_sampling_mode": "ddim",
    "ViTB32": true,
    "ViTB16": true,
    "ViTL14": true,
    "ViTL14_336px": false,
    "RN101": true,
    "RN50": true,
    "RN50x4": false,
    "RN50x16": false,
    "RN50x64": false,
    "ViTB32_laion2b_e16": false,
    "ViTB32_laion400m_e31": false,
    "ViTB32_laion400m_32": false,
    "ViTB32quickgelu_laion400m_e31": false,
    "ViTB32quickgelu_laion400m_e32": false,
    "ViTB16_laion400m_e31": false,
    "ViTB16_laion400m_e32": false,
    "RN50_yffcc15m": false,
    "RN50_cc12m": false,
    "RN50_quickgelu_yfcc15m": false,
    "RN50_quickgelu_cc12m": false,
    "RN101_yfcc15m": false,
    "RN101_quickgelu_yfcc15m": false,
    "cut_overview": "[10]*200+[8]*200+[6]*200+[4]*200+[2]*200",
    "cut_innercut": "[0]*200+[2]*200+[4]*200+[6]*200+[1]*200",
    "cut_ic_pow": "[0]*200+[10]*200+[10]*200+[10]*200+[1]*200",
    "cut_icgray_p": "[0.7]*100+[0.4]*100+[0.2]*100+[0.1]*100+[0]*600",
    "key_frames": true,
    "angle": "0:(0)",
    "zoom": "0: (1), 10: (1.05)",
    "translation_x": "0: (0)",
    "translation_y": "0: (0)",
    "translation_z": "0: (10.0)",
    "rotation_3d_x": "0: (0)",
    "rotation_3d_y": "0: (0)",
    "rotation_3d_z": "0: (0)",
    "midas_depth_model": "dpt_large",
    "midas_weight": 0.3,
    "near_plane": 200,
    "far_plane": 10000,
    "fov": 40,
    "padding_mode": "border",
    "sampling_mode": "bicubic",
    "video_init_path": "init.mp4",
    "extract_nth_frame": 2,
    "video_init_seed_continuity": false,
    "turbo_mode": false,
    "turbo_steps": "3",
    "turbo_preroll": 10,
    "use_horizontal_symmetry": false,
    "use_vertical_symmetry": false,
    "transformation_percent": [
        0.09
    ],
    "video_init_steps": 100,
    "video_init_clip_guidance_scale": 1000,
    "video_init_tv_scale": 0.1,
    "video_init_range_scale": 150,
    "video_init_sat_scale": 300,
    "video_init_cutn_batches": 4,
    "video_init_skip_steps": 50,
    "video_init_frames_scale": 15000,
    "video_init_frames_skip_steps": "70%",
    "video_init_flow_warp": true,
    "video_init_flow_blend": 0.999,
    "video_init_check_consistency": false,
    "video_init_blend_mode": "optical flow"
}

Example3.md

Example by WAS

Another fantastic, realistic and yet more complete portrait by WAS on the DD Discord's pwompt_land channel.

The prompt is four items long, showing that a simple prompt probably isn't the key variable in nailing coherence.

Full-size imae here

The full settings

{
    "text_prompts": {
        "0": [
            "Portrait of a Beautiful Fantasy Asian Woman in a Red Kimono, trending on ArtStation:5",
            "Photorealistic, Detailed, Realism, Fantasy, Vulemetric Lighting, Global Illumination, Subsurface Scattering:2",
            "Photographic Color Scheme",
            "blur, bokeh, dof:-1",
            "Red Light, Red Lighting, Red Environment:-2"
        ]
    },
    "image_prompts": {},
    "clip_guidance_scale": 2000,
    "tv_scale": 100,
    "range_scale": 150,
    "sat_scale": 0,
    "cutn_batches": 1,
    "max_frames": 10000,
    "interp_spline": "Linear",
    "init_image": null,
    "init_scale": 1000,
    "skip_steps": 10,
    "frames_scale": 1500,
    "frames_skip_steps": "60%",
    "perlin_init": false,
    "perlin_mode": "mixed",
    "skip_augs": false,
    "randomize_class": true,
    "clip_denoised": false,
    "clamp_grad": true,
    "clamp_max": 0.05,
    "seed": 3138,
    "fuzzy_prompt": false,
    "rand_mag": 0.05,
    "eta": 0.8,
    "width": 512,
    "height": 768,
    "diffusion_model": "custom",
    "use_secondary_model": false,
    "steps": 250,
    "diffusion_steps": 1000,
    "diffusion_sampling_mode": "ddim",
    "ViTB32": true,
    "ViTB16": false,
    "ViTL14": true,
    "ViTL14_336px": false,
    "RN101": false,
    "RN50": false,
    "RN50x4": true,
    "RN50x16": false,
    "RN50x64": false,
    "ViTB32_laion2b_e16": false,
    "ViTB32_laion400m_e31": false,
    "ViTB32_laion400m_32": false,
    "ViTB32quickgelu_laion400m_e31": false,
    "ViTB32quickgelu_laion400m_e32": false,
    "ViTB16_laion400m_e31": false,
    "ViTB16_laion400m_e32": false,
    "RN50_yffcc15m": false,
    "RN50_cc12m": false,
    "RN50_quickgelu_yfcc15m": false,
    "RN50_quickgelu_cc12m": false,
    "RN101_yfcc15m": false,
    "RN101_quickgelu_yfcc15m": false,
    "cut_overview": "[12]*400+[4]*600",
    "cut_innercut": "[4]*400+[12]*600",
    "cut_ic_pow": "[0.05]*1000",
    "cut_icgray_p": "[0.2]*400+[0]*600",
    "key_frames": true,
    "angle": "0:(0)",
    "zoom": "0: (1), 10: (1.05)",
    "translation_x": "0: (0)",
    "translation_y": "0: (0)",
    "translation_z": "0: (10.0)",
    "rotation_3d_x": "0: (0)",
    "rotation_3d_y": "0: (0)",
    "rotation_3d_z": "0: (0)",
    "midas_depth_model": "dpt_large",
    "midas_weight": 0.3,
    "near_plane": 200,
    "far_plane": 10000,
    "fov": 40,
    "padding_mode": "border",
    "sampling_mode": "bicubic",
    "video_init_path": "/content/drive/MyDrive/init.mp4",
    "extract_nth_frame": 2,
    "video_init_seed_continuity": false,
    "turbo_mode": false,
    "turbo_steps": "3",
    "turbo_preroll": 10,
    "use_horizontal_symmetry": false,
    "use_vertical_symmetry": false,
    "transformation_percent": [
        0.09
    ],
    "video_init_steps": 100,
    "video_init_clip_guidance_scale": 1000,
    "video_init_tv_scale": 0.1,
    "video_init_range_scale": 150,
    "video_init_sat_scale": 300,
    "video_init_cutn_batches": 4,
    "video_init_skip_steps": 50,
    "video_init_frames_scale": 15000,
    "video_init_frames_skip_steps": "70%",
    "video_init_flow_warp": true,
    "video_init_flow_blend": 0.999,
    "video_init_check_consistency": false,
    "video_init_blend_mode": "optical flow"
}