Bria @ SIGGRAPH 2025

Join Bria.ai at SIGGRAPH 2025 to explore the most developer-optimized, enterprise-ready visual generative AI platform on the market.

Vancouver Convention Centre | August 10–14, 2025 | Booth #725

What can I expect by visiting the Bria booth?

Join In Bria's Gen AI Coding Sessions

Tues August 12 - Thurs August 14

10 am - 5 pm

Tailored Fine‑Tuning: Brand‑Aligned Foundations

• Begin with Bria's source-available foundation models (via Hugging Face), which feed domain-specific datasets, such as product catalogs, brand imagery, or styled assets, to fine-tune custom variants (e.g., Bria‑4B‑Adapt).

• Showcase a pipeline that ingests brand-labeled training images and outputs a tailored model, preserving visual style, color palette, and composition of a brand.

Inpainting with Gen‑Fill - using an input image and a binary mask to insert or modify objects. For example, add a logo to a sneaker or place a new accessory on a product image.

• Blob‑optimized masks for smooth blending

• Precise ROI editing with full‑resolution output

• Optional content moderation safeguards

• APIs, iFrames, or ComfyUI nodes for easy no-code integration.

Smart Object Removal with Eraser: learn how to automate object removal (e.g., watermarks, blemishes) with seamless inpainting.

• Manual brush input and automatic mask generation

• Bria 2.3 backbone ensures high‑fidelity results and pixel‑perfect masking

• Compare against Canva/Firefly/StabilityAI benchmarking, highlighting Bria's superior artifact‑free erasures.
  

Product‑Shot Generation for eCommerce: Create automated pipelines for creating eye-catching product shots 

• Cutout: isolate items cleanly from the background

• Packshot: generate polished studio-like shots with solid or customizable backdrops

• Shadow & lifestyle shot: Add consistent shadows or contextual environments driven by text or image prompts.

• Demonstrate end-to-end flow: raw photo → cutout → packshot → lifestyle placement, all via API or iframe, yielding brand-consistent visuals across product catalogs.

Live Demo: 3D Character Skins and Character Shifts

Advanced image generation models can create high-quality, customizable skins and textures for 3D animated characters in real-time, streamlining the game development and animation pipeline

Step 1: Fine-Tune a "defined Character" Model

• Create a Fine-Tuning project (IP type: defined_character) via Bria's Console or API.

• Upload a curated dataset of 50–200 images of your character (e.g., mascot, avatar) in various poses and contexts.

• Automatically generate structured caption prefixes for each image—for example:

• Use LoRA-based fine-tuning (e.g., on BRIA-4B-Adapt) for fast, brand-aligned adaptation.

• Result: a lightweight, custom model that reliably reproduces your character's proportions, features, and style.

Step 2: Use ControlNet to Guide Skin Variations

• Choose a relevant ControlNet type based on your needs (e.g., Canny edges to preserve pose; Depth, Color-Grid to maintain structure/intensity).

• Upload a line-art or reference pose of your character.

• "Mascot in festive holiday outfit—red and gold"

• "Mascot wearing summer beach attire"

• "Local-themed mascot: festival costume with lanterns"

• Provide prompts like:

• At inference, ControlNet ensures that the pose, geometry, and silhouette remain fixed across variations.

• Fine-tune conditioning strength to strike a balance between creativity and identity retention.

Step 3: Automate Variant Prototype Workflows

• Reference Input → extract edges or depth maps via Canny/Depth ControlNet

• Prompt for style/theme (seasonal, regional, campaign-specific)

• Generate: Custom model + ControlNet → delivers variant skin

• Loop through multiple prompts and reference types to produce dozens of unique designs in minutes.

Step 4: Use-Cases & Benefits