What is Chroma1-HD?
Flux.1 [dev] was a great model, but it had two major issues for widespread community use.
- Since it is strongly distilled, compatibility with LoRA and full fine-tuning is not very good.
- Non-commercial license.
Therefore, several attempts appeared to reconstruct "Undistilled Flux" starting from Flux.1 [schnell] which has a more generous license, and one of the most labor-intensive ones among them is Chroma1-HD by LodestoneRock.
Since it has been re-trained including NSFW and anime images, it has a higher degree of freedom in style than the original Flux series, and it fits better to perceive it as a new general-purpose model based on Flux architecture rather than "Undistilled Flux".
Model Download
Unlike Flux.1, Chroma1-HD does not use CLIP, but uses only T5.
- diffusion_models
- text_encoders
- vae
📂ComfyUI/
└── 📂models/
├── 📂diffusion_models/
│ └── Chroma1-HD.safetensors
├── 📂text_encoders/
│ └── t5xxl_fp8_e4m3fn_scaled.safetensors
└── 📂vae/
└── ae.safetensors
text2image
{
"id": "a6c871b8-10c5-4758-ba11-0c13a6d67507",
"revision": 0,
"last_node_id": 57,
"last_link_id": 88,
"nodes": [
{
"id": 45,
"type": "CLIPTextEncode",
"pos": [
428.4686584472656,
402.1686706542969
],
"size": [
392.7272644042969,
107.36180877685547
],
"flags": {},
"order": 7,
"mode": 0,
"inputs": [
{
"name": "clip",
"type": "CLIP",
"link": 71
}
],
"outputs": [
{
"name": "CONDITIONING",
"type": "CONDITIONING",
"links": [
72
]
}
],
"properties": {
"cnr_id": "comfy-core",
"ver": "0.3.31",
"Node name for S&R": "CLIPTextEncode"
},
"widgets_values": [
"low quality, bad anatomy, extra digits, missing digits, extra limbs, missing limbs"
]
},
{
"id": 44,
"type": "T5TokenizerOptions",
"pos": [
153.3099365234375,
320.68304443359375
],
"size": [
210,
82
],
"flags": {},
"order": 5,
"mode": 0,
"inputs": [
{
"name": "clip",
"type": "CLIP",
"link": 70
}
],
"outputs": [
{
"name": "CLIP",
"type": "CLIP",
"links": [
69,
71
]
}
],
"properties": {
"cnr_id": "comfy-core",
"ver": "0.3.31",
"Node name for S&R": "T5TokenizerOptions"
},
"widgets_values": [
1,
0
],
"color": "#223",
"bgcolor": "#335"
},
{
"id": 42,
"type": "CLIPLoader",
"pos": [
-142.1746063232422,
320.68304443359375
],
"size": [
270,
106
],
"flags": {},
"order": 0,
"mode": 0,
"inputs": [],
"outputs": [
{
"name": "CLIP",
"type": "CLIP",
"links": [
70
]
}
],
"properties": {
"cnr_id": "comfy-core",
"ver": "0.3.31",
"Node name for S&R": "CLIPLoader"
},
"widgets_values": [
"t5xxl_fp8_e4m3fn.safetensors",
"chroma",
"default"
],
"color": "#432",
"bgcolor": "#653"
},
{
"id": 57,
"type": "MarkdownNote",
"pos": [
155.898193359375,
-99.75691986083984
],
"size": [
353.5063781738281,
220.58982849121094
],
"flags": {},
"order": 1,
"mode": 0,
"inputs": [],
"outputs": [],
"properties": {},
"widgets_values": [
"## models\n- [Chroma1-HD.safetensors](https://huggingface.co/lodestones/Chroma1-HD/blob/main/Chroma1-HD.safetensors)\n- [t5xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn_scaled.safetensors)\n- [ae.safetensors](https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors)\n\n\n```\n📂ComfyUI/\n└── 📂models/\n ├── 📂diffusion_models/\n │ └── Chroma1-HD.safetensors\n ├── 📂text_encoders/\n │ └── t5xxl_fp8_e4m3fn_scaled.safetensors\n └── 📂vae/\n └── ae.safetensors\n```"
],
"color": "#323",
"bgcolor": "#535"
},
{
"id": 8,
"type": "VAEDecode",
"pos": [
1247.5003662109375,
208.6058349609375
],
"size": [
156,
46
],
"flags": {},
"order": 9,
"mode": 0,
"inputs": [
{
"name": "samples",
"type": "LATENT",
"link": 52
},
{
"name": "vae",
"type": "VAE",
"link": 63
}
],
"outputs": [
{
"name": "IMAGE",
"type": "IMAGE",
"slot_index": 0,
"links": [
88
]
}
],
"properties": {
"cnr_id": "comfy-core",
"ver": "0.3.30",
"Node name for S&R": "VAEDecode"
},
"widgets_values": []
},
{
"id": 41,
"type": "VAELoader",
"pos": [
945,
86.43731689453125
],
"size": [
270,
58
],
"flags": {
"collapsed": false
},
"order": 2,
"mode": 0,
"inputs": [],
"outputs": [
{
"name": "VAE",
"type": "VAE",
"links": [
63
]
}
],
"properties": {
"cnr_id": "comfy-core",
"ver": "0.3.30",
"Node name for S&R": "VAELoader"
},
"widgets_values": [
"ae.safetensors"
],
"color": "#322",
"bgcolor": "#533"
},
{
"id": 56,
"type": "SaveImage",
"pos": [
1436.000732421875,
208.6058349609375
],
"size": [
421.36083984375,
467.84173583984375
],
"flags": {},
"order": 10,
"mode": 0,
"inputs": [
{
"name": "images",
"type": "IMAGE",
"link": 88
}
],
"outputs": [],
"properties": {
"cnr_id": "comfy-core",
"ver": "0.3.56"
},
"widgets_values": [
"ComfyUI"
]
},
{
"id": 49,
"type": "EmptySD3LatentImage",
"pos": [
555.8366088867188,
575.5158081054688
],
"size": [
261.60076904296875,
106
],
"flags": {},
"order": 3,
"mode": 0,
"inputs": [],
"outputs": [
{
"name": "LATENT",
"type": "LATENT",
"slot_index": 0,
"links": [
75
]
}
],
"properties": {
"cnr_id": "comfy-core",
"ver": "0.3.39",
"Node name for S&R": "EmptySD3LatentImage"
},
"widgets_values": [
1152,
1152,
1
]
},
{
"id": 31,
"type": "KSampler",
"pos": [
900,
208.6058349609375
],
"size": [
315,
262
],
"flags": {},
"order": 8,
"mode": 0,
"inputs": [
{
"name": "model",
"type": "MODEL",
"link": 87
},
{
"name": "positive",
"type": "CONDITIONING",
"link": 67
},
{
"name": "negative",
"type": "CONDITIONING",
"link": 72
},
{
"name": "latent_image",
"type": "LATENT",
"link": 75
}
],
"outputs": [
{
"name": "LATENT",
"type": "LATENT",
"slot_index": 0,
"links": [
52
]
}
],
"properties": {
"cnr_id": "comfy-core",
"ver": "0.3.30",
"Node name for S&R": "KSampler"
},
"widgets_values": [
777,
"fixed",
26,
6,
"euler",
"beta",
1
]
},
{
"id": 55,
"type": "UNETLoader",
"pos": [
546.31689453125,
38.45512008666992
],
"size": [
270,
82
],
"flags": {},
"order": 4,
"mode": 0,
"inputs": [],
"outputs": [
{
"name": "MODEL",
"type": "MODEL",
"links": [
87
]
}
],
"properties": {
"cnr_id": "comfy-core",
"ver": "0.3.56",
"Node name for S&R": "UNETLoader"
},
"widgets_values": [
"Chroma\\Chroma1-HD.safetensors",
"default"
],
"color": "#323",
"bgcolor": "#535"
},
{
"id": 43,
"type": "CLIPTextEncode",
"pos": [
421.19598388671875,
196.53317260742188
],
"size": [
400,
149.17999267578125
],
"flags": {},
"order": 6,
"mode": 0,
"inputs": [
{
"name": "clip",
"type": "CLIP",
"link": 69
}
],
"outputs": [
{
"name": "CONDITIONING",
"type": "CONDITIONING",
"links": [
67
]
}
],
"properties": {
"cnr_id": "comfy-core",
"ver": "0.3.31",
"Node name for S&R": "CLIPTextEncode"
},
"widgets_values": [
"A stylish young lady with short, wavy white hair, rendered in a stylish anime or manga art style.textile shading. She has piercing, light-colored eyes with sharp eyelashes, and a confident, slightly seductive gaze looking directly at the viewer. His lips are slightly parted. He is wearing a dark, sleeveless top. His left hand is raised to his face, fingers elegantly posed, with long, black painted nails and a gold ring on his index finger. He appears to be holding a cigarette near his lips, with smoke subtly depicted."
]
}
],
"links": [
[
52,
31,
0,
8,
0,
"LATENT"
],
[
63,
41,
0,
8,
1,
"VAE"
],
[
67,
43,
0,
31,
1,
"CONDITIONING"
],
[
69,
44,
0,
43,
0,
"CLIP"
],
[
70,
42,
0,
44,
0,
"CLIP"
],
[
71,
44,
0,
45,
0,
"CLIP"
],
[
72,
45,
0,
31,
2,
"CONDITIONING"
],
[
75,
49,
0,
31,
3,
"LATENT"
],
[
87,
55,
0,
31,
0,
"MODEL"
],
[
88,
8,
0,
56,
0,
"IMAGE"
]
],
"groups": [],
"config": {},
"extra": {
"ds": {
"scale": 0.8264462809917354,
"offset": [
181.6746063232422,
242.10691986083992
]
},
"frontendVersion": "1.35.0",
"VHS_latentpreview": false,
"VHS_latentpreviewrate": 0,
"VHS_MetadataImage": true,
"VHS_KeepIntermediate": true
},
"version": 0.4
}
- The basic construction is almost the same as Flux.1.
- 🟦
T5TokenizerOptionsis a node that fills with padding when the number of tokens is less than the set value. There is no big difference whether you put it in or not. - CFG is specified as 4.0, but some say 6 to 7 is good. Please try various things.