AI Briefing#text-to-image
The latest AI and developer news about #text-to-image, with the original source and a short summary.
The latest AI and developer news about #text-to-image, with the original source and a short summary.
Qwen-Image-Flash demonstrates the importance of optimizing data composition and training pipelines, beyond objective functions, for accelerating image generation models.
Ideogram has unveiled **Ideogram 4**, its first open-weight text-to-image generation model.
MONET, an open-source dataset consisting of over 100 million high-quality images and captions, has been released.
Qwen has released Q-Judger, a VLM model that automatically evaluates text-to-image generation quality.
MAI-Image-2.5 has been released, ranking 3rd on the Arena text-to-image leaderboard.
PrismML has unveiled Bonsai Image 4B, a 1-bit/ternary-based model that can run locally in the browser via WebGPU.
NTM aims for both 4-step generation and exact likelihood by training the reverse step as a normalizing flow.
Artificial Analysis ranked the anonymous model Peanut #8 on its Text-to-Image leaderboard.
By combining ElevenLabs' voice AI with FLUX Kontext, the author built Vibe Draw, which generates and edits images through speech.
Photoroom used 32 H200 GPUs to release an optimized recipe and code for training a text-to-image model in just 24 hours.
Recraft V4, a text-to-image model for professional design, has launched in AI Gateway.
Qwen-Image-2.0 integrates PPTs, posters, and photos into a single model with 1k-token instructions and 2K resolution.
Photoroom has released experimental analysis and optimization techniques for efficient Text-to-Image model training.
Qwen-Image-2512 significantly improves human realism, natural detail, and text rendering.
By finding material embeddings in CLIP-space, it implements material blending and fine-grained attribute control.
The Hugging Face community has released an open-source preference dataset for training text-to-image generation models.
Implemented natural likeness using an HG-DPO-based backbone model and the Instantbooth module, validated via a photo booth at if(kakaoAI)2024.
A leaderboard comparing the performance of image generation models based on over 45,000 human preferences has launched.
Improved noise sampling for Rectified Flow models and introduced a new Transformer architecture to enhance high-resolution image synthesis performance.
aMUSEd, an efficient non-diffusion text-to-image generation model that recreates Google's MUSE, has been released.
This explains how to efficiently deploy the DeepFloyd IF model, which has a multi-stage architecture, using BentoML.
This analyzes the causes of bias occurring in text-to-image (TTI) models and emphasizes the need for improved methodologies to evaluate it.
rinna has released 'Japanese Stable Diffusion,' which reflects Japanese prompts and Japanese culture.