Generation of furniture images from photos and sketches (Stable Diffusion / ControlNet / LoRA)

AI & Machine Learning — incorrectly specified categories?

752 USD

Project translated automatically. Log in or register, to view the original

A custom furniture image generation system (folding beds, modules, dressers, and other products) needs to be created based on provided photographs and sketches.
The goal is to obtain realistic visualizations of furniture in interiors with maximum accuracy in preserving the geometry, proportions, and materials of the original products.

Training data:
– 10–15 photographs of furniture from different angles;
– sketches or elevation drawings (JPG/PNG);
– a selection of textures and examples of interiors (if necessary).

System requirements:
– training a custom model based on LoRA or DreamBooth using the provided data;
– configuring ControlNet (Canny, Depth, LineArt, etc.) for accurate preservation of shapes and angles;
– the ability to generate based on text descriptions (changing facades, colors, backgrounds, interiors, etc.);
– integration into AUTOMATIC1111 WebUI or ComfyUI with ready-made presets;
– providing 3–5 demonstration images confirming the quality of the result.

Work results:
– trained model (.safetensors / checkpoint) with version control;
– configurations, examples of prompts, and a set of ControlNet models with descriptions;
– detailed instructions for running locally and in Colab;
– complete transfer of all files and rights without restrictions on use.

Deadlines and budget:
– completion time — up to 10 working days;
– estimated budget — 45,000 ₽ (negotiable);
– work format — remotely, entirely in Russian.

Additionally:
The test task is ready — I will provide a set of photos and sketches immediately after agreement.
If clarifications on technical details or data formats are needed, I am ready to respond promptly.

Proposals 2 Discussions 1

Artur Onysko

0 0

Projects -
Rating -
Rating 187

Budget: 660 EUR Deadline: 10 days

My name is Arthur, I am a professional engaged in the implementation and training of neural network solutions for visualization and image generation for specific tasks.
I am ready to deliver your project "turnkey" — from data collection and annotation to creating a custom model and integration into AUTOMATIC1111 / ComfyUI.

What I offer:
• Training a custom model on your photos and sketches (DreamBooth / LoRA).
• Configuration and integration of ControlNet (Canny, Depth, LineArt, etc.) for accurate reproduction of the shape and materials of furniture.
• Generation of photorealistic images of furniture in an interior based on a text description (replacement of facades, colors, backgrounds, etc.).
• Support and consultation at every stage — from setup to final result.
• Detailed instructions for running locally and in Colab, as well as a complete package of files for independent work.

I am ready to complete a test task and show the result within 6–10 working days.
I explain everything as simply as possible, I can connect and guide you through the entire process personally.
I work for results — until the project is fully delivered.

I am ready to start immediately after receiving the initial data.
I am waiting for confirmation and your materials to start!

The list does not show proposals concealed by the client or freelancer with a Plus profile, as well as proposals violating rules

Aleksei Sergeev
Castelln de la Plana, Spain

Projects -
Rating -
Rating 40