The model mainly trains two categories:

film photography and computational photography

, and thoroughly fine-tunes Asian portraits.

Add some popular film models while optimizing core data.

Images are

rated for image quality, negative features are extracted from various low-quality images, and image quality control capabilities

are enhanced

.

Face optimization LORA is now online, and three types of faces have been added. Welcome to taste: https://www.liblib.ai/modelinfo/737f163de6014b3da545c8f6133dc0d1?mine=1 This model

uses As the basic model, Leosam's Hello World 5.0 model (HW5.0 for short) is the basic model, and all usage rules also follow the HW5.0 statement. If you make fusion modifications to this model, be sure to mention this model and HW5.0 in the introduction.

The following is the HW5.0 model address: https://www.liblib.ai/modelinfo/506c46c91b294710940bd4b183f3ecd7 The data subject uses GPT4V marking, and some use WD1.4+CogVqA. The XL model is not easy to train. Every training and tuning will take a lot of time, effort, and computing power. Please abide by the

copyright statement.

If you like my model, please buy me a cup of coffee or support me with iPower Generation to speed up model training and research! We also welcome more feedback and criticism. This is really important to me! If you have any business cooperation, please contact us: SDSDCC33 https://afdian.net/a/SDSDCC?tab=home

Overview of training reminders: For resolution, 896*1152

is recommended. For other parameters, please refer to V4RC1 or HW5.0

instructions.

The following are key training concepts and summoning films: Film photography Some film models

: Fuji

C100 shooting, Fuji C200 shooting, Kodak 400

Shooting, Kodak Gold 200 shooting, Nolan 5219 shooting Computational Photography: Computational Photography Quality Rating: Mobile

Phone image quality, Landline image quality, Pager image quality

(image quality in descending order: mobile image quality, landline image quality, pager image quality) Trained negative hints: overexposed background, poor lighting, overexposed areas, difficult

Lighting, low resolution, potential compression artifact (I tested this part and doesn't interfere too much with the picture) Usage recommendations: When using film, enter

Film photography+ any film (or without film name)

using negative tips from training and worst quality, low quality, illustration, 3d, 2d, painting, cartoons, sketch, optional,

plus

computational Photography, mobile phone image quality, Landline image quality, and Pager image quality completely embedding negative

computational photography categories may cause the risk of excessive sharpening. At the same time, since the data subject is still film, it is generally a film

effect by default without adding specific reminders.

When using computational photography (that is, mobile image quality), entering computational photography+any negative image quality rating does not include negative training hints,

because these are also characteristics of computational photography.

Comparison of image quality in different photography categories:

Comparison of different film models:

Comparison of different negative embeddings:

Note: None of the photography effects represent the effects of film or mobile phone shooting in the real world. These are all AI simulation results, and also have personal aesthetics. Please do not map the model effects to specific devices in reality.