@lobehub/chat
Version:
Lobe Chat - an open-source, high-performance chatbot framework that supports speech synthesis, multimodal, and extensible Function Call plugin system. Supports one-click free deployment of your private ChatGPT/LLM web application.
336 lines (319 loc) • 11.9 kB
text/typescript
import { ModelParamsSchema, PRESET_ASPECT_RATIOS } from '../standard-parameters';
import { AIImageModelCard } from '../types';
/**
* Aspect ratios supported by FLUX models
* Support wide range ratios from 21:9 to 9:21, including foldable screen devices
*/
const FLUX_ASPECT_RATIOS = [
'21:9', // Ultra-wide screen
'16:9', // Widescreen
'8:7', // Foldable screen (e.g. Galaxy Z Fold, unfolded state ~7.6 inch)
'4:3', // Traditional landscape
'3:2', // Classic landscape
'1:1', // Square
'2:3', // Classic portrait
'3:4', // Traditional portrait
'7:8', // Foldable screen portrait
'9:16', // Portrait
'9:21', // Ultra-tall portrait
];
/**
* Standard aspect ratios supported by SD models
* Based on preset aspect ratios, suitable for traditional SD model use cases
*/
const SD_ASPECT_RATIOS = PRESET_ASPECT_RATIOS;
/**
* Extended aspect ratios supported by SDXL models
* Support more modern display ratios, similar to FLUX but more conservative
*/
const SDXL_ASPECT_RATIOS = [
'16:9', // Modern widescreen
'4:3', // Traditional landscape
'3:2', // Classic landscape
'1:1', // Square
'2:3', // Classic portrait
'3:4', // Traditional portrait
'9:16', // Modern portrait
];
/**
* FLUX.1 Schnell model parameter configuration
* Ultra-fast text-to-image mode, generates in 1-4 steps, Apache 2.0 license
*/
export const fluxSchnellParamsSchema: ModelParamsSchema = {
aspectRatio: {
default: '1:1',
enum: FLUX_ASPECT_RATIOS,
},
cfg: { default: 1, max: 1, min: 1, step: 0 }, // Schnell uses fixed CFG of 1
height: { default: 1024, max: 1536, min: 512, step: 8 },
prompt: { default: '' },
samplerName: { default: 'euler' },
scheduler: { default: 'simple' },
seed: { default: null },
steps: { default: 4, max: 4, min: 1, step: 1 },
width: { default: 1024, max: 1536, min: 512, step: 8 },
};
/**
* FLUX.1 Dev model parameter configuration
* High-quality text-to-image mode, supports guidance scale adjustment, non-commercial license
*/
export const fluxDevParamsSchema: ModelParamsSchema = {
aspectRatio: {
default: '1:1',
enum: FLUX_ASPECT_RATIOS,
},
cfg: { default: 3.5, max: 10, min: 1, step: 0.5 },
height: { default: 1024, max: 2048, min: 512, step: 8 },
prompt: { default: '' },
samplerName: { default: 'euler' },
scheduler: { default: 'simple' },
seed: { default: null },
steps: { default: 20, max: 50, min: 10, step: 1 },
width: { default: 1024, max: 2048, min: 512, step: 8 },
};
/**
* FLUX.1 Krea-dev model parameter configuration
* Enhanced safety text-to-image mode, developed in collaboration with Krea, non-commercial license
*/
export const fluxKreaDevParamsSchema: ModelParamsSchema = {
aspectRatio: {
default: '1:1',
enum: FLUX_ASPECT_RATIOS,
},
cfg: { default: 3.5, max: 10, min: 1, step: 0.5 },
height: { default: 1024, max: 2048, min: 512, step: 8 },
prompt: { default: '' },
samplerName: { default: 'dpmpp_2m_sde' },
scheduler: { default: 'karras' },
seed: { default: null },
steps: { default: 15, max: 50, min: 10, step: 1 },
width: { default: 1024, max: 2048, min: 512, step: 8 },
};
/**
* FLUX.1 Kontext-dev model parameter configuration
* Image editing mode, supports modifying existing images based on text instructions, non-commercial license
*/
export const fluxKontextDevParamsSchema: ModelParamsSchema = {
cfg: { default: 3.5, max: 10, min: 1, step: 0.5 },
imageUrl: { default: '' }, // Input image URL (supports text-to-image and image-to-image)
prompt: { default: '' },
seed: { default: null },
steps: { default: 28, max: 50, min: 10, step: 1 }, // Kontext defaults to 28 steps
strength: { default: 0.85, max: 1, min: 0, step: 0.05 }, // Image editing strength control (frontend parameter)
};
/**
* SD3.5 model parameter configuration
* Stable Diffusion 3.5, supports Large and Medium versions, automatically selects by priority
*/
export const sd35ParamsSchema: ModelParamsSchema = {
aspectRatio: {
default: '1:1',
enum: FLUX_ASPECT_RATIOS, // SD3.5 also supports multiple aspect ratios
},
cfg: { default: 4, max: 20, min: 1, step: 0.5 },
height: { default: 1024, max: 2048, min: 512, step: 8 },
prompt: { default: '' },
samplerName: { default: 'euler' },
scheduler: { default: 'sgm_uniform' },
seed: { default: null },
steps: { default: 20, max: 50, min: 10, step: 1 },
width: { default: 1024, max: 2048, min: 512, step: 8 },
};
/**
* SD1.5 text-to-image model parameter configuration
* Stable Diffusion 1.5 text-to-image generation, suitable for 512x512 resolution
*/
export const sd15T2iParamsSchema: ModelParamsSchema = {
aspectRatio: {
default: '1:1',
enum: SD_ASPECT_RATIOS,
},
cfg: { default: 7, max: 20, min: 1, step: 0.5 },
height: { default: 512, max: 1024, min: 256, step: 8 },
prompt: { default: '' },
samplerName: { default: 'euler' },
scheduler: { default: 'normal' },
seed: { default: null },
steps: { default: 25, max: 50, min: 10, step: 1 },
width: { default: 512, max: 1024, min: 256, step: 8 },
};
/**
* SDXL text-to-image model parameter configuration
* SDXL text-to-image generation, suitable for 1024x1024 resolution
*/
export const sdxlT2iParamsSchema: ModelParamsSchema = {
aspectRatio: {
default: '1:1',
enum: SDXL_ASPECT_RATIOS,
},
cfg: { default: 8, max: 20, min: 1, step: 0.5 },
height: { default: 1024, max: 2048, min: 512, step: 8 },
prompt: { default: '' },
samplerName: { default: 'euler' },
scheduler: { default: 'normal' },
seed: { default: null },
steps: { default: 30, max: 50, min: 10, step: 1 },
width: { default: 1024, max: 2048, min: 512, step: 8 },
};
/**
* SDXL image-to-image model parameter configuration
* SDXL image-to-image generation, supports input image modification
*/
export const sdxlI2iParamsSchema: ModelParamsSchema = {
cfg: { default: 8, max: 20, min: 1, step: 0.5 },
imageUrl: { default: '' }, // Input image URL
prompt: { default: '' },
samplerName: { default: 'euler' },
scheduler: { default: 'normal' },
seed: { default: null },
steps: { default: 30, max: 50, min: 10, step: 1 },
strength: { default: 0.75, max: 1, min: 0, step: 0.05 }, // Image modification strength (frontend parameter)
};
/**
* Custom SD text-to-image model parameter configuration
* Custom Stable Diffusion text-to-image model with flexible parameter settings
*/
export const customSdT2iParamsSchema: ModelParamsSchema = {
aspectRatio: {
default: '1:1',
enum: SDXL_ASPECT_RATIOS, // Use broader aspect ratio support
},
cfg: { default: 7, max: 30, min: 1, step: 0.5 },
height: { default: 768, max: 2048, min: 256, step: 8 },
prompt: { default: '' },
samplerName: { default: 'euler' }, // Use SDXL common parameters
scheduler: { default: 'normal' }, // Use SDXL common parameters
seed: { default: null },
steps: { default: 25, max: 100, min: 5, step: 1 },
width: { default: 768, max: 2048, min: 256, step: 8 },
};
/**
* Custom SD image-to-image model parameter configuration
* Custom Stable Diffusion image-to-image model, supports image editing
*/
export const customSdI2iParamsSchema: ModelParamsSchema = {
cfg: { default: 7, max: 30, min: 1, step: 0.5 },
imageUrl: { default: '' }, // Input image URL
prompt: { default: '' },
samplerName: { default: 'euler' }, // Use SDXL common parameters
scheduler: { default: 'normal' }, // Use SDXL common parameters
seed: { default: null },
steps: { default: 25, max: 100, min: 5, step: 1 },
strength: { default: 0.75, max: 1, min: 0, step: 0.05 }, // Image modification strength (frontend parameter)
};
/**
* List of image generation models supported by ComfyUI
* Supports FLUX series and Stable Diffusion 3.5 models
*/
const comfyuiImageModels: AIImageModelCard[] = [
{
description:
'FLUX.1 Schnell - 超快速文生图模型，1-4步即可生成高质量图像，适合实时应用和快速原型制作',
displayName: 'FLUX.1 Schnell',
enabled: true,
id: 'comfyui/flux-schnell',
parameters: fluxSchnellParamsSchema,
releasedAt: '2024-08-01',
type: 'image',
},
{
description: 'FLUX.1 Dev - 高质量文生图模型，10-50步生成，适合高质量创作和艺术作品生成',
displayName: 'FLUX.1 Dev',
enabled: true,
id: 'comfyui/flux-dev',
parameters: fluxDevParamsSchema,
releasedAt: '2024-08-01',
type: 'image',
},
{
description: 'FLUX.1 Krea-dev - 增强安全的文生图模型，与 Krea 合作开发，内置安全过滤',
displayName: 'FLUX.1 Krea-dev',
enabled: false,
id: 'comfyui/flux-krea-dev',
parameters: fluxKreaDevParamsSchema,
releasedAt: '2025-07-31',
type: 'image',
},
{
description:
'FLUX.1 Kontext-dev - 图像编辑模型，支持基于文本指令修改现有图像，支持局部修改和风格迁移',
displayName: 'FLUX.1 Kontext-dev',
enabled: true,
id: 'comfyui/flux-kontext-dev',
parameters: fluxKontextDevParamsSchema,
releasedAt: '2025-05-29', // Aligned with BFL official Kontext series release date
type: 'image',
},
{
description:
'Stable Diffusion 3.5 新一代文生图模型，支持 Large 和 Medium 两个版本，需要外部 CLIP 编码器文件，提供卓越的图像质量和提示词匹配度。',
displayName: 'Stable Diffusion 3.5',
enabled: true,
id: 'comfyui/stable-diffusion-35',
parameters: sd35ParamsSchema,
releasedAt: '2024-10-22',
type: 'image',
},
{
description:
'Stable Diffusion 3.5 内置 CLIP/T5 编码器版本，无需外部编码器文件，适用于 sd3.5_medium_incl_clips 等模型，资源占用更少。',
displayName: 'Stable Diffusion 3.5 (内置编码器)',
enabled: false,
id: 'comfyui/stable-diffusion-35-inclclip',
parameters: sd35ParamsSchema,
releasedAt: '2024-10-22',
type: 'image',
},
{
description:
'Stable Diffusion 1.5 文生图模型，经典的512x512分辨率文本到图像生成，适合快速原型和创意实验',
displayName: 'SD 1.5',
enabled: false,
id: 'comfyui/stable-diffusion-15',
parameters: sd15T2iParamsSchema,
releasedAt: '2022-08-22',
type: 'image',
},
{
description:
'SDXL 文生图模型，支持1024x1024高分辨率文本到图像生成，提供更好的图像质量和细节表现',
displayName: 'SDXL 文生图',
enabled: true,
id: 'comfyui/stable-diffusion-xl',
parameters: sdxlT2iParamsSchema,
releasedAt: '2023-07-26',
type: 'image',
},
{
description:
'SDXL 图生图模型，基于输入图像进行高质量的图像到图像转换，支持风格迁移、图像修复和创意变换。',
displayName: 'SDXL Refiner',
enabled: true,
id: 'comfyui/stable-diffusion-refiner',
parameters: sdxlI2iParamsSchema,
releasedAt: '2023-07-26',
type: 'image',
},
{
description:
'自定义 SD 文生图模型，模型文件名请使用 custom_sd_lobe.safetensors，如有 VAE 请使用 custom_sd_vae_lobe.safetensors，模型文件需要按照 Comfy 的要求放入对应文件夹',
displayName: '自定义 SD 文生图',
enabled: false,
id: 'comfyui/stable-diffusion-custom',
parameters: customSdT2iParamsSchema,
releasedAt: '2023-01-01',
type: 'image',
},
{
description:
'自定义 SDXL 图生图模型，模型文件名请使用 custom_sd_lobe.safetensors，如有 VAE 请使用 custom_sd_vae_lobe.safetensors，模型文件需要按照 Comfy 的要求放入对应文件夹',
displayName: '自定义 SDXL Refiner',
enabled: false,
id: 'comfyui/stable-diffusion-custom-refiner',
parameters: customSdI2iParamsSchema,
releasedAt: '2023-01-01',
type: 'image',
},
];
export const allModels = [...comfyuiImageModels];
export default allModels;