### Session: ION-CROSS

Expected run time

170 minutes

Just the Instagram channel name.

We do not currently support private channels.

This feature is new. We'd love your feedback!

Enter a distinctive name to serve as the trigger that activates your LORA model.

This name acts as a unique identifier, allowing you to summon your model with a specific word or phrase in any prompt.

Select the base model you would like to train a LoRA for.

The base model selection will impact runtime, price, and available options.

Pro Tip: Select "Data set only" to make sure the training data is what you want before training the LoRA.

Check this box to save the dataset we train your LoRA with.

You can re-use a saved dataset later by selecting it under the "Show more options" section on the "Video to LoRA" page.

Generate 4 sample images using your new LoRA with custom generated prompts based on the imagery found in the training data.

Adjust or disable face selection.

Use a reference image if your subject's face is not the most frequent on their channel.

Face detection is enabled by default. It is the magic behind our character LoRA process.

Leave face detection enabled unless you are trying to make a style or concept LoRA.

Face detection will remove images that do not contain your subject.

Face detection will crop face and body shots from all images that match and that are big enough (at least 300px wide or tall).

Lower values mean more similarity and fewer matches. Below 20 nearly all shots of the same person will not match.

Higher values mean more tolerance and more matches. Above 80 most different people will match each other.

If your subject's face is the most frequently shown, we can select it for you automatically.

Even if your subject is not the most frequently shown in any one video, as long as it is more common than any other face across all source videos for this job, we will find it.

To specify a specific face, upload a reference image.

This option makes the dataset NSFW prior to training.

Your original images will be digitally altered to remove the clothing from the subject in the images.

A LoRA trained on this NSFW dataset will generate high quality NSFW images more consistently.

The dataset can be fun to look at, too.

### Create a LoRA (Low Rank Adaptation)

model from an Instagram channel

#### Instagram to LoRA

- WAN 2.2 T2V (24 tokens)
- WAN 2.1 T2V (24 tokens)
- Hunyuan Video (24 tokens)
- Qwen Image (24 tokens)
- Z-Image Turbo (19 tokens)
- FLUX (19 tokens)
- SDXL (9 tokens)
- Data set only (8 tokens)

We will extract up to 500 images from the source media. A minimum of 25 images is required. When face detection is enabled, we also require a minimum of 20 faces of the same human that are at least 300 pixels wide or tall. Jobs that fail will be refunded or resubmitted automatically. If that does not happen quickly enough, contact us through our form or on Discord.

### Instagram Channel

**Trigger**

Show more options

- Enable Face Detection
- Exclude if different by more than: 60%
- Auto-Detect Most Common Face
- Upload Reference Image

***Select files:***

Reference

[ PNG, JPG ]

Save training data set (+1 token)

Generate sample images (+3 tokens)

Make dataset NSFW (+50 tokens)

**22 tokens** Train LoRA

Uploading video to One Shot server...