Custom LoRA training — your character, your model

Send me your photos, I train a LoRA optimized for the Enzino ecosystem. Backed by 110K+ downloads and a perfect 5.0 rating on Civitai, this is expert-level training, not a generic service.

How it works

  1. Send your images
  2. Dataset preparation
  3. Training
  4. Your LoRA is ready

Pricing

Custom Lora — $99+

  • Please provide the following information with your order
  • What would you like to train?
  • Briefly describe your character, person, style or concept
  • Dataset
  • Please provide your images in the best quality possible

Custom Character — $149+

  • Your own unique AI character, created from scratch
  • Tell me what you have in mind - appearance, age, hair, style, personality, aesthetic - and I'll design a consistent, realistic character exclusively for you
  • Once the character is approved, I'll create the image dataset and train a dedicated LoRA so you can generate your character whenever you want
  • Original character creation
  • Character design & visual identity

Celebrity Lora — $49+

  • Want a specific celebrity available as a high-quality LoRA?
  • Choose the public figure you'd like to see - actress, singer, model or other well-known personality - and I'll create a dedicated LoRA optimized for realistic generations
  • Celebrity of your choice
  • LoRA training
  • Testing & optimization

Dataset requirements

  • Minimum 15 high-quality photos (more = better)
  • Variety of angles: front, 3/4, profile, slight up/down
  • Variety of lighting: natural, indoor, outdoor, different times
  • Variety of expressions: neutral, smile, serious, etc.
  • No filters, no Snapchat, no heavy makeup changes
  • No sunglasses or accessories covering the face
  • Minimum resolution: 1024x1024 or equivalent
  • Consistent appearance (same hair, no drastic changes)

Working with me

  • Your photos, your LoRA — You provide the dataset. I don't generate or source reference photos.
  • 100% upfront — Payment before work starts. No deposits, no installments, no exceptions.
  • AI is not pixel-perfect — Outputs will vary between generations. That's how AI works — not a bug.
  • Post-processing matters — FaceDetailer, upscale, and other tools may alter the LoRA's base output. This is normal.
  • Limited revisions — Each tier includes a set number of rounds. Major scope changes = new project.
  • Support via ticket only — No Discord DMs, no email threads. Tickets keep things clean and documented.
  • Right to refuse — I reserve the right to stop a project if the client is abusive or provides unusable material.
  • 5–14 business days — Turnaround depends on complexity. Rush delivery available on request.

Frequently asked questions

Why is payment 100% upfront?

Custom LoRA training involves significant compute time and manual work before any output exists. Upfront payment ensures commitment from both sides and lets me focus on quality, not chasing invoices.

What if my dataset gets rejected?

I'll tell you exactly what's wrong and what to fix. You can resubmit as many times as needed before we proceed. I'd rather delay than train on bad data.

How many revisions are included?

Standard includes 1 round, Advanced includes 2. An adjustment means tweaking the LoRA based on your feedback (face too round, skin tone off, etc.). Major scope changes count as a new project.

Can I use the LoRA commercially?

Not by default. Commercial rights are available as an add-on to any tier. Contact me for details and pricing based on your use case.

Why can't I get support on Discord?

To maintain quality and accountability, all support goes through the ticket system. It keeps everything documented, organized, and ensures nothing falls through the cracks. Discord is for community and announcements only.

What's the difference between Standard and Advanced?

Standard covers most use cases: dataset cleaning, captioning, training, and 1 adjustment round. Advanced adds extended training iterations, edge-case testing across multiple prompts, and 2 adjustment rounds. Pick Advanced if you need rock-solid consistency or plan heavy use.

Do you create characters or generate the dataset?

No. You provide the photos, I train the LoRA. I handle cleaning, captioning, and optimization — but the source material is yours. This ensures better results and protects both of us.

What base models do you support?

Training is optimized for the Enzino ecosystem — primarily IntoRealism and ZIT. If you have a specific base model requirement, mention it in your project submission and I'll let you know if it's supported.