Door categorieën bladeren
Ontdekken
Fiverr Pro
Nederlands
$
USD
I will train a lora for your consistent ai influencer, character or mascot
Maxim M
Over deze dienst
Get the same character in every image. I train a custom LoRA for Stable Diffusion XL on my own NVIDIA H100, so your AI character, mascot or product keeps its face, colors and details in any scene you prompt.
What you get
- A LoRA file (.safetensors) ready for ComfyUI, Automatic1111 and Forge, plus a diffusers copy
- Test images of your character in new scenes, with the prompts used
- A ready ComfyUI workflow in the Standard and Premium packages
- A trigger word and simple prompting tips
How I do it
- I curate and caption the training images: an inconsistent set is the main reason character LoRAs fail
- Only 1-3 reference images? The Premium package first builds a consistent training set (image-prompt generation, similarity scoring, manual review), then trains the LoRA
- Fixed prompts and seeds with and without the LoRA, so you see what it changes
Recent result: an original robot character reproduced in new scenes (surfing, cooking, cycling) from a curated set of 21 images, trained in 17 minutes.
Message me your reference images and style before ordering, and I will confirm the package. Real people only with their consent.
Voorkeur voor een leveringsstijl
Laat het de freelancer weten als je voorkeuren voor of zorgen hebt over het gebruik van AI-tools voor het voltooien en/of leveren van je bestelling.
Verklaring AI-avatar design
Ik begrijp dat ik, door diensten voor AI-avatar design aan te kopen, alle disclaimerregels moet volgen die vereist zijn door de toepasselijke wetgeving en door de platform(s) waar deze content geplaatst wordt. Lees meer over de AI-richtlijnen van Fiverr.
Maak kennis met Maxim M
Maxim M
- Afkomstig uitRusland
- Lid sindsmrt 2025
Talen
Russisch, Engels
I’m an AI and Machine Learning specialist with a strong focus on data annotation, neural network training, and data analysis. With hands-on experience in building and deploying machine learning models, I help businesses leverage artificial intelligence to solve complex challenges. Whether you need accurate data labeling, custom AI solutions, or insights from large datasets, I can provide expert support tailored to your needs.
My Expertise Includes:
Data Annotation: Labeling and preparing data for supervised learning tasks (image, text, and audio).
Machine Learning & AI Models: Designing etc
Mijn portfolio
Veelgestelde vragen
How many images do I need?
15-30 clear images of the character from different angles for Basic and Standard. For Premium, 1-3 good reference images are enough: I build the training set from them.
Which models do you support?
SDXL and SDXL-based checkpoints, for example realistic or anime fine-tunes. For Flux or SD 1.5, message me first and I will confirm.
Will the character look exactly the same?
A good LoRA keeps the main identity (face, colors, shapes) across scenes, while small details can vary between seeds. I show fixed-seed comparisons so you can judge before you accept.
Can I use the images commercially?
Yes, for images you generate with your LoRA, as long as you own the rights to the reference images. I do not train on real people without their consent.
Do I need ComfyUI?
No. The LoRA works in ComfyUI, Automatic1111, Forge and diffusers. The ComfyUI workflow is included so you can start generating right away.
