# text-to-image A machine-learning technology that generates images from text prompts. This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles. ## Introducing ChatGPT Images 2.5 DevFeed: [Introducing ChatGPT Images 2.5]() Original publisher: [Read original article]() Author: Simon Willison Published: 2026-09-08T22:46:33Z Content type: release Language: en Sources: [Simon Willison's Weblog]() Topics: [ChatGPT](), [OpenAI](), [Generative AI](), [text-to-image](), [Command-line interface]() Tags: [ai](), [ai-2-236](), [chatgpt](), [generative-ai](), [generative-ai-1-982](), [openai](), [openai-463](), [text-to-image](), [text-to-image-47](), [tools](), [tools-78](), [uv](), [uv-100]() ### AI overview OpenAI's ChatGPT Images 2.5 release adds improved multi-turn instruction following, faster responses, and better preservation of subjects in reference photos. The API introduces the gpt-image-2.5-sunburst and gpt-image-2.5-flare model IDs, with Sunburst positioned for precise editing and Flare for faster everyday generation. ### Source excerpt Introducing ChatGPT Images 2.5 OpenAI's image generation models are apparently used "more than 3 billion images across ChatGPT Images and the GPT-Image models in the API". This latest release improves their instruction-following ability across multiple turns, responds faster, and "is better at preserving the subjects in your reference photos". There are two new model IDs in the API: gpt-image-2.5-sunburst and gpt-image-2.5-flare. Based on this I think Sunburst is the stronger option: Choose Sunburst for workflows where editing precision matters most, and Flare for fast, high-quality everyday image generation. I upgraded my openai_image.py CLI tool to support passing in one or more reference images, so now this works: uv run https://tools.simonwillison.net/python/openai_image.py \ 'add a raccoon scientist studying the chart thoughtfully' \ -i https://static.simonwillison.net/static/2026/openai-agent-usage.webp \ -m gpt-image-2.5-sunburst This is the original image, and here's what I got back from that prompt to "add a raccoon scientist studying the chart thoughtfully": Tags: tools, ai, openai, generative-ai, uv, text-to-image ## Alice AI ART 2.0: путь к unified-модели, которая одинаково хорошо умеет генерировать и редактировать картинки DevFeed: [Alice AI ART 2.0: путь к unified-модели, которая одинаково хорошо умеет генерировать и редактировать картинки]() Original publisher: [Read original article]() Author: snk4tr (Яндекс) Published: 2026-07-14T08:40:58Z Content type: article Language: ru Sources: [Яндекс - Как мы делаем Яндекс / Статьи]() Topics: [Artificial Intelligence](), [text-to-image]() Tags: [ai](), [ai-art](), [alice-ai](), [image-editing](), [image-to-image](), [tag-5db67948c7d9](), [tag-d346fb5ae499](), [tag-e493b9551702](), [text-to-image](), [unified-models]() ### AI overview A team developing Alice AI ART describes its effort to unify text-to-image generation and instruction-based image editing in one model. The article covers the previous split architecture, the motivation for unification, and the evaluation datasets used to measure quality. ### Source excerpt Привет, Хабр! На связи команда генеративных моделей в компьютерном зрении. Вместе с другими командами мы делаем мультимодального ассистента Алиса AI. Внутри него мы развиваем несколько вариантов визуальной генерации с помощью отдельной модели Alice AI ART. Два базовых сценария её работы -- генерация по тексту (Text-to-Image, T2I) и редактирование по картинке с инструкцией (Image-to-Image, I2I). Именно о них пойдёт речь. Всё это время эти сценарии жили как два разных стека: свои базовые модели, свои данные, свои метрики и, честно говоря, своя отдельная боль в разработке и поддержке. В этом году мы поставили себе цель, которая звучала просто, а на практике оказалась полугодовым приключением: не только подтянуть качество, а сделать одну модель, которая одинаково хорошо умеет и в T2I, и в I2I. Внутри мы называем такой режим unified или просто uni. Вас ждёт рассказ об отдельных экспериментах и наблюдениях, которые помогли нам сделать первый шаг в этом направлении и привели нас к Alice AI ART 2.0, -- включая те, которые красиво не сработали (спойлер: их хватало). Читать далее ## Start building with Nano Banana 2 Lite and Gemini Omni Flash DevFeed: [Start building with Nano Banana 2 Lite and Gemini Omni Flash]() Original publisher: [Read original article]() Author: Alisa Fortin Published: 2026-06-30T16:02:40Z Content type: release Language: en Sources: [Google DeepMind News]() Topics: [Google AI](), [API](), [text-to-image](), [Artificial Intelligence]() Tags: [ai](), [api](), [benchmarks](), [build-faster](), [building](), [cost](), [efficiency](), [gemini](), [generation](), [generative](), [google-ai](), [images](), [latency](), [life](), [media](), [models](), [none](), [performance](), [prototyping](), [scale](), [video]() ### AI overview Google DeepMind announces Nano Banana 2 Lite, a fast and cost-efficient image model for high-throughput developer pipelines, and makes Gemini Omni Flash available to developers for video generation and conversational editing. The models are offered through Google AI Studio, the Gemini API and Gemini Enterprise Agent Platform. ### Source excerpt Scale your ideas with Nano Banana 2 Lite, our fastest, most cost-efficient Gemini Image model, and Gemini Omni Flash for high-quality video and conversational editing. ## DeepInfra on Hugging Face Inference Providers 🔥 DevFeed: [DeepInfra on Hugging Face Inference Providers 🔥]() Original publisher: [Read original article]() Author: Aray Sultanbekova; Shang-Pin; Utemuratov; Yessen K; Oguz Vuruskaner; Célina Hanouti; Simon Brandeis; Lucain Pouget Published: 2026-04-29T00:00:00Z Content type: article Language: en Sources: [Hugging Face - Blog]() Topics: [deepinfra](), [hugging face](), [inference-providers](), [AI Inference](), [SDKs](), [Large Language Model](), [API keys](), [text-generation](), [text-to-image]() Tags: [ai-inference](), [api](), [api-keys](), [applications](), [artificial-intelligence](), [code](), [deepinfra](), [enterprise](), [hub](), [hugging-face](), [inference](), [inference-providers](), [javascript](), [llms](), [open](), [open-source](), [partnerships](), [pricing](), [python](), [sdks](), [text-generation](), [text-to-image]() ### AI overview DeepInfra is now a supported Inference Provider on the Hugging Face Hub, with serverless access to more than 100 models and integration with Hugging Face's JavaScript and Python SDKs. The article describes provider selection, API-key and routed-by-Hugging-Face modes, supported model tasks, and initial access to conversational and text-generation models. ### Source excerpt We're on a journey to advance and democratize artificial intelligence through open source and open science. ## PRX Part 3 -- Training a Text-to-Image Model in 24h! DevFeed: [PRX Part 3 -- Training a Text-to-Image Model in 24h!]() Original publisher: [Read original article]() Author: David Bertoin; Roman Frigg; Jon Almazán Published: 2026-03-03T16:50:49Z Content type: article Language: en Sources: [Hugging Face - Blog]() Topics: [text-to-image](), [GPU](), [Code](), [Hardware]() Tags: [code](), [compute](), [computer-vision](), [cost](), [diffusion](), [engineering](), [experiment](), [gpu](), [hardware](), [model](), [open](), [performance](), [text-to-image](), [training]() ### AI overview The article describes a 24-hour speedrun for training a text-to-image diffusion model by combining architectural and training techniques from earlier posts. It targets 512px training followed by 1024px fine-tuning on 32 H200 GPUs within an approximately $1,500 compute budget, and says the training code and experimental framework will be open-sourced. ### Source excerpt Welcome back 👋 In the last two posts (Part 1 and Part 2), we explored a wide range of architectural and training tricks for diffusion models. We tried to evaluate each idea in isolation, measuring throughput, convergence speed, and final image quality, and tried to understand what actually moves the needle. In this post, we want to answer a much more practical question: What happens when we combine all the tricks that worked? ## Training Design for Text-to-Image Models: Lessons from Ablations DevFeed: [Training Design for Text-to-Image Models: Lessons from Ablations]() Original publisher: [Read original article]() Author: David Bertoin; Roman Frigg; Jon Almazán Published: 2026-02-03T11:25:53Z Content type: article Language: en Sources: [Hugging Face - Blog]() Topics: [text-to-image](), [foundation-models](), [Optimization]() Tags: [efficiency](), [experiments](), [models](), [open](), [optimization](), [scale](), [series](), [techniques](), [text-to-image](), [training]() ### AI overview This second article in a series documents experiments for training efficient text-to-image foundation models from scratch. Using PRX as the case study, it examines techniques that improve training speed, convergence reliability, and learned representations, comparing them through controlled experiments and combinations against a clean Flow Matching baseline. ### Source excerpt Welcome back! This is the second part of our series on training efficient text-to-image models from scratch. In the first post of this series, we introduced our goal: training a competitive text-to-image foundation model entirely from scratch, in the open, and at scale. We focused primarily on architectural choices and motivated the core design decisions behind our model PRX. ## Build with Nano Banana Pro, our Gemini 3 Pro Image model DevFeed: [Build with Nano Banana Pro, our Gemini 3 Pro Image model]() Original publisher: [Read original article]() Author: Alisa Fortin Published: 2025-11-20T15:11:14Z Content type: article Language: en Sources: [Google DeepMind News]() Topics: [Google AI](), [text-to-image](), [Artificial Intelligence](), [API](), [multimodal](), [Google Search](), [Figma]() Tags: [4k](), [ai](), [ai-studio](), [api](), [color-grading](), [demo](), [figma](), [gemini](), [generation](), [google-search](), [image](), [model](), [multimodal](), [none](), [product-design]() ### AI overview The article announces Nano Banana Pro, also called Gemini 3 Pro Image, a higher-fidelity image generation and editing model built on Gemini 3 Pro. It describes developer access through the Gemini API in Google AI Studio and Vertex AI, with multimodal capabilities, Google Search grounding, improved text rendering, precise image controls, high-resolution outputs, and support for compositing multiple reference elements. The model is also being integrated into developer and creative platforms for UI mockups and visual asset creation. ### Source excerpt Nano Banana Pro, or Gemini 3 Pro Image, is our most advanced image generation and editing model. ## Replicate Integration with Encore.ts DevFeed: [Replicate Integration with Encore.ts]() Original publisher: [Read original article]() Author: Ivan Cernja Published: 2025-11-17T00:00:00Z Content type: tutorial Language: en Sources: [Encore Updates]() Topics: [AI Development](), [text-to-image](), [API](), [flux](), [stable-diffusion](), [Node.js](), [Secrets Management]() Tags: [ai](), [api](), [diffusion](), [flux](), [integration](), [node-js](), [secrets-management](), [stable-diffusion](), [tutorial]() ### AI overview This tutorial explains how to build a type-safe backend with Encore.ts and Replicate to run AI model predictions and generate images. It covers asynchronous results, image storage, API authentication, secret management, and models such as FLUX and Stable Diffusion. ### Source excerpt Running AI models with Replicate and Encore ## DigitalOcean Gradient AI Platform adds image generation, knowledge-base auto-indexing, VPC security, and developer tools DevFeed: [DigitalOcean Gradient AI Platform adds image generation, knowledge-base auto-indexing, VPC security, and developer tools]() Original publisher: [Read original article]() Author: Grace Morgan Published: 2025-10-02T07:12:50Z Content type: release Language: en Sources: [DigitalOcean]() Topics: [AI Platform](), [Artificial Intelligence](), [Digital Ocean](), [text-to-image](), [API](), [Serverless](), [stable-diffusion](), [multimodal](), [Security](), [Image](), [OpenAI]() Tags: [2025](), [accelerate](), [ai](), [ai-platform](), [ai-tools](), [api](), [api-integration](), [authentication](), [code](), [data](), [development](), [generation](), [github](), [image](), [openai](), [platform](), [product-updates](), [range](), [text-to-image](), [vpc](), [workflow]() ### AI overview DigitalOcean announces new Gradient AI Platform capabilities, including programmatic image generation through the Serverless Inference API, Knowledge Base Auto-Indexing, security controls, and additional developer tools for building AI applications. ### Source excerpt At Deploy London 2025, we shared the next chapter of the Gradient AI Platform. We're making it easier for developers and businesses to build production-ready AI applications, whether you're experimenting with your first agent or scaling an enterprise workload. Today, in that spirit, we're introducing a new wave of features that expand what you can build with the Gradient AI Platform, give you greater security and control, and accelerate your development workflow. Expanding capabilities for Gradient AI Platform AI applications are becoming more multimodal and data-driven, able to work with text, images, audio, and other formats. With Image Model Support and Knowledge Base Auto-Indexing, these new Gradient AI Platform features make it easier than ever to give your agents a wide range of inputs and knowledge sources. Image Model Support You can now generate images programmatically using text prompts through Gradient AI Platform's Serverless Inference API, powered by OpenAI's gpt-image-1 model. This is the platform's first non-text modality (with more coming soon), expanding our capabilities from text-only to include text-to-image generation. Text-to-image generation - Generate images directly via the Serverless Inference API. API integration - OpenAI compatible using the image generation endpoint. For code examples and setup instructions, check out the DigitalOcean Gradient AI starter kit on GitHub. Unified billing - Image generation charges appear on your DigitalOcean account alongside other services. Developer-first workflow - Direct API access without requiring UI-based tools or self-hosting. Enterprise authentication - Use Model Access Keys for secure access. With this capability, you can create images using natural language prompts for projects such as content generation, marketing assets, product imagery, and more. Get started with Image Model Support -> Knowledge Base Auto-Indexing Keep your agents up to date without the manual work. With auto-indexing, new and u ## Fast LoRA inference for Flux with Diffusers and PEFT DevFeed: [Fast LoRA inference for Flux with Diffusers and PEFT]() Original publisher: [Read original article]() Author: Sayak Paul; Benjamin Bossan Published: 2025-07-23T00:00:00Z Content type: article Language: en Sources: [Hugging Face - Blog]() Topics: [flux](), [lora](), [Inference Performance](), [Hacktoberfest](), [diffusers](), [Optimization](), [peft](), [quantization](), [model architecture](), [text-to-image](), [GPU]() Tags: [amd](), [diffusers](), [diffusion](), [flux](), [gpu](), [guide](), [inference](), [latency](), [lora](), [model-architecture](), [optimization](), [peft](), [quantization](), [text-to-image]() ### AI overview This article presents an optimization recipe for faster LoRA inference with the Flux.1-Dev text-to-image model using Diffusers and PEFT. The approach addresses LoRA hotswapping and recompilation issues with Flash Attention 3, FP8 quantization from TorchAO, and hotswapping-ready compilation, achieving about a 2.3x speedup while balancing inference speed and memory use. ### Source excerpt We're on a journey to advance and democratize artificial intelligence through open source and open science. ## Memory-efficient Diffusion Transformers with Quanto and Diffusers DevFeed: [Memory-efficient Diffusion Transformers with Quanto and Diffusers]() Original publisher: [Read original article]() Author: Sayak Paul; David Corvoysier Published: 2024-07-30T00:00:00Z Content type: tutorial Language: en Sources: [Hugging Face - Blog]() Topics: [diffusers](), [diffusion-transformers](), [quantization](), [optimum](), [PyTorch](), [stable-diffusion](), [text-to-image](), [benchmarking]() Tags: [benchmarking](), [diffusers](), [diffusion-transformers](), [efficiency](), [gpu](), [guide](), [optimum](), [pytorch](), [quality](), [quantization](), [sd3](), [text-to-image]() ### AI overview This tutorial explains how to reduce the memory requirements of Transformer-based diffusion pipelines using Quanto quantization utilities from the Diffusers library. It benchmarks FP8 quantization on PixArt-Sigma, Stable Diffusion 3, and Aura Flow, reporting memory savings with slightly higher latency and little quality degradation. ### Source excerpt We're on a journey to advance and democratize artificial intelligence through open source and open science. ## AI and Apple Intelligence: Apple AI features, dog-bark decoding, Mistral, Stable Diffusion, and ChatGPT learning DevFeed: [AI and Apple Intelligence: Apple AI features, dog-bark decoding, Mistral, Stable Diffusion, and ChatGPT learning]() Original publisher: [Read original article]() Author: Pradeep Kumar Published: 2024-06-19T14:32:53Z Content type: news Language: en Sources: [Air Around AI]() Topics: [Artificial Intelligence](), [genai](), [AI Models](), [text-to-image](), [Open Source](), [ChatGPT](), [datasets]() Tags: [ai](), [apple](), [artificial-intelligence](), [chatgpt](), [datasets](), [diffusion](), [open-source]() ### AI overview This news roundup covers Apple Intelligence features for Siri, image and speech recognition, privacy-focused on-device processing, AI research on dog vocalizations, Mistral AI's reported $6 billion valuation, Stable Diffusion 3 Medium, Apple's release of Core ML models and datasets on Hugging Face, and a ChatGPT-based learning example. ### Source excerpt Siri with AI, Decoding Dog barks, Mistral is now $6 Billion, Apple's Open 20 Core ML Model & 4 datasets, 80/20 Principle with ChatGPT, Gartner's super 3 predictions for GenAI, Dream Machine ## Diffusers welcomes Stable Diffusion 3 DevFeed: [Diffusers welcomes Stable Diffusion 3]() Original publisher: [Read original article]() Author: Dhruv Nair; YiYi Xu; Sayak Paul; Alvaro Somoza; Kashif Rasul; Apolinário from multimodal AI art Published: 2024-06-12T00:00:00Z Content type: release Language: en Sources: [Hugging Face - Blog]() Topics: [stable-diffusion](), [diffusers](), [diffusion-transformers](), [Artificial Intelligence](), [text-to-image](), [multimodal](), [lora](), [Inference](), [Open Source]() Tags: [artificial-intelligence](), [diffusers](), [diffusion](), [guide](), [inference](), [lora](), [multimodal](), [open-source](), [sd3](), [stable-diffusion](), [text-to-image]() ### AI overview The article announces Stable Diffusion 3 Medium, a 2B-parameter latent diffusion model. It describes the MMDiT architecture, multimodal text and image processing, rectified flow matching, inference support through a new scheduler, and accompanying Diffusers, DreamBooth, and LoRA resources. ### Source excerpt We're on a journey to advance and democratize artificial intelligence through open source and open science. ## Launching the Artificial Analysis Text to Image Leaderboard & Arena DevFeed: [Launching the Artificial Analysis Text to Image Leaderboard & Arena]() Original publisher: [Read original article]() Author: Micah Hill-Smith; George Cameron Published: 2024-06-06T00:00:00Z Content type: article Language: en Sources: [Hugging Face - Blog]() Topics: [text-to-image](), [Human-AI evaluation](), [Artificial Intelligence](), [Open Source](), [stable-diffusion]() Tags: [artificial-intelligence](), [collaboration](), [community](), [diffusion](), [evaluation](), [generation](), [leaderboard](), [open-source](), [openai](), [ranking](), [research](), [text-to-image]() ### AI overview The Artificial Analysis Text to Image Leaderboard ranks open-source and proprietary image-generation models using ELO scores derived from more than 45,000 human image preferences. Its Image Arena uses crowdsourced comparisons across diverse prompts and use cases. The article highlights the rapid evolution of image models, the strong performance of proprietary systems, and the growing competitiveness of open-source models such as Playground AI v2.5 and Stable Diffusion 3. ### Source excerpt We're on a journey to advance and democratize artificial intelligence through open source and open science. ## MobileDiffusion: Rapid text-to-image generation on-device DevFeed: [MobileDiffusion: Rapid text-to-image generation on-device]() Original publisher: [Read original article]() Author: Google AI (noreply@blogger.com) Published: 2024-01-31T21:59:00Z Content type: article Language: en Sources: [Google Research]() Topics: [Machine learning](), [Mobile](), [text-to-image](), [Inference](), [model-deployment](), [Optimization](), [Android](), [iOS](), [MediaPipe]() Tags: [android](), [computer-vision](), [deployment](), [diffusion](), [generation](), [image](), [inference](), [ios](), [machine-learning](), [mediapipe](), [mobile](), [on-device](), [on-device-learning](), [optimization](), [text-to-image]() ### AI overview Google researchers introduce MobileDiffusion, a latent diffusion model designed for on-device text-to-image generation. Using DiffusionGAN for one-step sampling, it was tested on premium iOS and Android devices and generated 512x512 images in about half a second with a 520-million-parameter model. ### Source excerpt Posted by Yang Zhao, Senior Software Engineer, and Tingbo Hou, Senior Staff Software Engineer, Core ML Text-to-image diffusion models have shown exceptional capabilities in generating high-quality images from text prompts. However, leading models feature billions of parameters and are consequently expensive to run, requiring powerful desktops or servers (e.g., Stable Diffusion, DALL-E, and Imagen). While recent advancements in inference solutions on Android via MediaPipe and iOS via Core ML have been made in the past year, rapid (sub-second) text-to-image generation on mobile devices has remained out of reach. To that end, in "MobileDiffusion: Subsecond Text-to-Image Generation on Mobile Devices", we introduce a novel approach with the potential for rapid text-to-image generation on-device. MobileDiffusion is an efficient latent diffusion model specifically designed for mobile devices. We also adopt DiffusionGAN to achieve one-step sampling during inference, which fine-tunes a pre-trained diffusion model while leveraging a GAN to model the denoising step. We have tested MobileDiffusion on iOS and Android premium devices, and it can run in half a second to generate a 512x512 high-quality image. Its comparably small model size of just 520M parameters makes it uniquely suited for mobile deployment. Rapid text-to-image generation on-device. Background The relative inefficiency of text-to-image diffusion models arises from two primary challenges. First, the inherent design of diffusion models requires iterative denoising to generate images, necessitating multiple evaluations of the model. Second, the complexity of the network architecture in text-to-image diffusion models involves a substantial number of parameters, regularly reaching into the billions and resulting in computationally expensive evaluations. As a result, despite the potential benefits of deploying generative models on mobile devices, such as enhancing user experience and addressing emerging privacy concer