https://www.youtube.com/watch?v=hbh6qnLE49I&feature=youtu.be

⭐️ Basics

Before going into the how, let’s talk about the why. Here are the 4 most common questions we receive from people joining Argil :

• Can you train images without creating your model ? (spoiler : yes)
• Why should you train a model ?
• How long does a training take ?
• Can we only train and generate images ? (spoiler : no)

⚙️ Training your first model

If you’re looking to train images on your specific use cases, such as people or objects, you need to “teach” them to the AI. This process is simple :

1. Go to “Models”

2. Click “Train a new model”

3. Follow the steps

   → Pick a name

   → Upload pictures (at least 10 for good results) of a person, object or anything visual you wish to replicate

   → Select the type of thing you’re training so we can define the right settings

   → Click “train” and receive an estimation of training time, then wait for your email

Once trained, you can access the model in the studio.

<aside> 💡 Training can take up to 1 hour depending on settings, so make sure to anticipate or launch several trainings at the same time!

</aside>

⚠️ Tips for picking the images to upload:

• It is preferable to train at least 10-15 good images of a subject in png or jpg format
• It’s preferable not to input bad images (bad quality, posture, colours) as it might be overrepresented in the results. 15 perfect photos is better than 20 photos with 5 bad apples.
• If human, 70-80% of the images should be closeup shots of the face