> ## Knowledge Base Index
> Fetch the complete knowledge base index at: https://help.seelab.ai/sitemap.xml
> Use this file to discover available pages before exploring further.
> Pure-Markdown content can be obtained by appending a '.md' suffix to the content URLs listed in the sitemap (without the trailing slash).

# Create and train a model

Creating a Seelab model takes a few minutes: you choose the type, import your images, mark the most representative ones and start the training. The model is then available to your whole project, in image and in video.

# Before you start

* **Prepare 1 to 8 images** of your subject, sharp, varied and at least 1024 px. See [Preparing training images](https://help.seelab.ai/en/article/how-to-photoshoot-for-seelab-ai-custom-training-1q7r2xc/).
* **Check the active project** at the top right: the model will be shared with the members of this project.

# Create the model in five steps

1. **Open Custom models** in the navigation bar, then click **Create new model**. For your first model, the type choice appears right away.
2. **Choose the type**: **Train your Avatar** (Portraits & Avatars), **Train your Product** (Products & Objects) or **Train your Style** (Styles & Photography).

![The three types of models](https://storage.crisp.chat/users/helpdesk/website/-/6/8/0/3/6803ea00be5f5800/image_3k90dw.jpg =600xauto)

3. **Name the model and import your images** in the **Training images** area, by drag and drop or from your computer.
4. **Mark the 3 key images** with the 👍 icon, then add a brief if needed with **Import a brief**: a PDF document and/or a text.

![The completed creation form](https://storage.crisp.chat/users/helpdesk/website/-/6/8/0/3/6803ea00be5f5800/image_1f3ojh8.jpg =1000xauto)

5. **Click Start training.** The **Remaining trainings** counter, at the top of the form, shows how many trainings you have left.

# During training

The model switches to **In training** status. Training takes from a few seconds to a few minutes, then Seelab automatically redirects you to your model.

![A model in training](https://storage.crisp.chat/users/helpdesk/website/-/6/8/0/3/6803ea00be5f5800/image_15h5ext.jpg =600xauto)

|| A training can be rejected if the images don't comply with the terms of use, or if their quality or consistency isn't sufficient. In that case, review your dataset or contact the Seelab team.

# What's next?

* **Your model is ready in image and in video**: a single training creates both.
* **It appears in the Custom models library**, with the **Use model** and **Edit** buttons.
* **The Trainings panel** on the model page shows your quotas: **Remaining Actives** (the number of models you can still train) and **Remaining Monthlies** (the retrainings left this month).
* **Every project member** finds it in their model selector.

To use it: [Using your model in image and video](https://help.seelab.ai/en/article/using-your-model-in-image-and-video-15ag1ou/).

# Tips

* **Test your model as soon as training ends** on three or four very different scenes: packshot, in context, close-up.
* **If a detail is missing** (label, material), add images that show it and run the training again with **Retrain** from the model page.
* **The Seelab team can handle it with you**: **Contact us** in the Custom models space, or the chat.

# Glossary

* **Remaining trainings**: the number of trainings you have left.
* **In training**: the status of a model that is being trained.
* **Retrain**: run the training of an existing model again with new images.
* **Key images**: the 3 images marked 👍, which best define the expected look.