Diffusion models with sufficiently large parameter counts can successfully process and render highly detailed textual scene descriptions

Overview Debates Related claims Topics

Canva Craft executes design generation across three distinct stages: prompt amplification, structured scene graph generation, and pixel synthesis

Canva trained Canva Craft on a dataset composed of stock imagery, synthetic data, user-permitted content, and millions of proprietary Canva designs

Canva maintains approximately 100 distinct artificial intelligence models operating in production

Canva announced Canva Craft, a foundational artificial intelligence model for image generation

Initial evaluations of Canva Craft on Design Arena revealed that generated cartoon images were excessively photorealistic and three-dimensional

Canva Craft uses structured, text-based scene descriptions containing bounding boxes, coordinates, and relative element positions

Canva Craft was trained entirely from scratch rather than fine-tuned from an existing open-weights base model

Canva Craft converts user prompts into structured scene graphs prior to pixel generation, allowing layout edits before final rendering

Training or fine-tuning specialized artificial intelligence models on proprietary domain data can surpass larger frontier models while utilizing two orders of magnitude fewer parameters

Claude Sonnet 4.5 was priced at 3 dollars per million input tokens and 15 dollars per million output tokens

OpenAI users generate approximately 4 billion images per week

According to Menlo Ventures, frontier models' share of enterprise AI spend rose from 35% to 63%

OpenAI updated its image generation model to version 2.6, adding support for generating animated GIFs

GPT-6 Astro is priced at 10 dollars per million input tokens and 60 dollars per million output tokens

GPT-5 was priced at 1.25 dollars per million input tokens and 10 dollars per million output tokens

Claude Opus 5.5 is priced at 4 dollars per million input tokens and 20 dollars per million output tokens

OpenAI's Images 2.6 model is 50% faster than its predecessor and can generate an image in as little as six seconds

Nicheformer is a spatial omics foundation model that predicts neighboring cell gene expression states based on an individual cell's transcriptomic profile

Fabian Theis's group uploaded a 100-million-parameter single-cell foundation model trained on 350 million cells across 16 species to Hugging Face

Fabian Theis's group applied self-supervised contrastive learning to develop the single-cell foundation model scMoMa