--- title: Custom Models slug: custom-models description: Learn how to train a custom model using Snowpixel and prompt it to generate content with this comprehensive document. Discover three different types of models: Object Model, Human Model, and Style Model. The easy-to-follow instructions outline the process docTags: createdAt: 2023-04-30T07:11:11.000Z --- ::::WorkflowBlock :::WorkflowBlockItem You can train your own custom model on your custom data. Upload a few images and in 20 minutes, your personal model will be ready. Now you can prompt anything from your newly trained model. You can train a Object, Human or Style Model. - Object Model. Use this if you want to train an Object like your pet or your favourite childhood toy. You will be able to prompt your childhood toy in any scenario. - Style Model. Use this if you want to train a particular style. Then you will be able to prompt anything in a particular style. - Human Model. It is a special type of Object Model which has some optimizations to make it work better on humans. Here is the workflow on how you can train a Custom Model and how to prompt it. ::: :::WorkflowBlockItem ### Go to [Snowpixel](https://make.snowpixel.app/) ::: :::WorkflowBlockItem ### Click on Train your own model button ![](https://api.archbee.com/api/optimize/F12aM8jg2IPyOMpKvHZ_d/Qx1BCAfDwxTVuHoQ1CdsG_ab-capture-image.png) ::: :::WorkflowBlockItem ### Give a name to your model We are creating model of character Wednesday Addams of the same named series, here give a name to your model. We are calling it "Wednesday" ![](https://api.archbee.com/api/optimize/F12aM8jg2IPyOMpKvHZ_d/UkO_dkmIKCTzZJsEM1nKi_ab-capture-image.png) ::: :::WorkflowBlockItem ### Chose what to train As Wednesday Addams is a human, we used train **Human** button ![](https://api.archbee.com/api/optimize/F12aM8jg2IPyOMpKvHZ_d/UqDBIbC3MobQsvfs4Kp4g_ab-capture-image.png) ::: :::WorkflowBlockItem ### Click on Upload Images and Train button First you're going to want to gather all of your images. To get the best results you'll want **ATLEAST 10 IMAGES** all of which contains the following : - The subject is in focus - Subject is in different context and lighting across all of the photos - Photos should only containing one person/thing or style that you want to train the model on - For photos that you have on your phone it's a good idea to use the crop tool to make sure everything looks good before uploading You can upload multiple images in various dimensions (landscape, portrait, square) and formats (jpg, png etc) by clicking on Upload Images and Train button. Note: Images not in square dimensions are automatically cropped in square. ![](https://api.archbee.com/api/optimize/F12aM8jg2IPyOMpKvHZ_d/Pd1kh9xbVARK0cxHUKx0E_ab-capture-image.png) Below are all the images we used to train model ![](https://api.archbee.com/api/optimize/F12aM8jg2IPyOMpKvHZ_d/he6OuOQeDedCAaelZD2oY_wa-dataset.png "Images used for training") ::: :::WorkflowBlockItem ### Click on Okay After images are uploaded you will get a pop-up, Click OK and wait for the email that you will receive when the custom model is finetuned. It will take between 5 to 20 minutes. If you have not received in this time, make sure to check your spam folder. ![](https://api.archbee.com/api/optimize/F12aM8jg2IPyOMpKvHZ_d/D09NZQNBoyvKXFpYJLBPq_ab-capture-image.png) ::: :::WorkflowBlockItem ### Click on Personal Model To use your trained model and generate content go to personal models ![](https://api.archbee.com/api/optimize/F12aM8jg2IPyOMpKvHZ_d/BCR1xUk0MJA3SqD1IqBua_ab-capture-image.png) ::: :::WorkflowBlockItem ### Click on Choose model dropdown ![](https://api.archbee.com/api/optimize/F12aM8jg2IPyOMpKvHZ_d/fZKntvglBwy6leqwDDSJR_1111.png) ::: :::WorkflowBlockItem ### Click on your model name "Wednesday" ![](https://api.archbee.com/api/optimize/F12aM8jg2IPyOMpKvHZ_d/rdUADhB3zGCNTIQkwVrNd_ab-capture-image.png) ::: :::WorkflowBlockItem ### Type "X on mount everest" Here "X" refers to Wednesday Addams. Whichever model you train, **"X" is a special keyword which refers to object/style/human of that model**. If you trained a style model, you would use prompt "Car in X style". There, "X" will refer to your specific style. ![](https://api.archbee.com/api/optimize/F12aM8jg2IPyOMpKvHZ_d/9MCWPTM3mViTKj2eJMUqe_ab-capture-image.png) ::: :::WorkflowBlockItem ### Click on Generate Make sure your custom model is selected from dropdown here before proceeding ![](https://api.archbee.com/api/optimize/F12aM8jg2IPyOMpKvHZ_d/NaN0KlGkKFumkG8IfEKKR_ab-capture-image.png) ::: :::WorkflowBlockItem ### You will see In progress tab You will also receive an email once generations are completed and they will also be automatically visible on this page once they are generated ![](https://api.archbee.com/api/optimize/F12aM8jg2IPyOMpKvHZ_d/v0oKauY3RCo8Xt3t5U39C_ab-capture-image.png) ::: :::WorkflowBlockItem ### Click on Next page button Check out all images by Clicking on Next page button at the bottom of screen ![](https://api.archbee.com/api/optimize/F12aM8jg2IPyOMpKvHZ_d/-p5boaAK59D-t-ztgxI_5_ab-capture-image.png) ::: :::WorkflowBlockItem ### Click on download You can use Download all images button at the top or get selected images by clicking on download button under any image ![](https://api.archbee.com/api/optimize/F12aM8jg2IPyOMpKvHZ_d/CCrvHuOw5qt1PzlZZdYDF_ab-capture-image.png) ::: :::WorkflowBlockItem ### Click on OK After this you will receive a mail with download link of the image(s). This will take a minute or two as your images are being upscaled to 2k resolution ![](https://api.archbee.com/api/optimize/F12aM8jg2IPyOMpKvHZ_d/4w5YWOPgTPzz_9I-ZBDp3_ab-capture-image.png) ::: :::WorkflowBlockItem ### Here are some of the results ![](https://api.archbee.com/api/optimize/F12aM8jg2IPyOMpKvHZ_d/5pyzPMjdZD4U4lK0OYLbN_wa-generations.png "Generated images from custom model") ::: ::::