Replicate
AI Infra
Cloud ML inference platform. Run open-source models via API.
Recent activity
-
Grok Imagine Video 1.5 is the most exciting video model release from xAI. You can generate realistic video with synchronized audio in a single pass, capable of juggling complex motion with precise prompt adherence. We pushed it hard across a range of scenes, and came up with the ultimate prompting guide to get the most out of this model.
Read more → -
Seedance 2.0 is a revolutionary AI video model that can create high-quality, cinematic videos from a combination of up to 9 images, 3 video clips, 3 audio files, and a text prompt, allowing users to replicate complex scenes and sequences. AI summary
Read more → -
Seedream 5.0 is a powerful image model that understands photographic language at a deep level, replicating the look and feel of various film stocks, lenses, and lighting setups with high accuracy. It can also be used to edit images by referencing before/after pairs and applying transformations to new scenes. The model's ability to reason through prompts and understand physical objects at a mechanical level allows for logical and physically correct results. AI summary
Read more → -
Recraft V4 generates art-directed images — and actual editable SVGs — with strong composition, accurate text rendering, and what the Recraft team calls "design taste." Four models are available on Replicate now.
Read more → -
Isaac 0.1, a 2B-parameter open-weight vision-language model, has been released on Replicate, allowing developers to try it out with a JavaScript API. The model excels in grounded perception, visual reasoning, OCR, and spatial awareness, making it suitable for applications requiring transparency and real-time processing. Isaac can learn new tasks from examples without fine-tuning and is efficient enough for real-time or edge-constrained applications. AI summary
Read more → -
The article announces the release of FLUX.2, a highly advanced image generation model by Black Forest Labs, offering significant improvements in image quality, editing capabilities, and enterprise-grade efficiency. FLUX.2 is available in three variants: FLUX.2 [pro], FLUX.2 [flex], and FLUX.2 [dev], each with varying levels of quality and features. AI summary
Read more → -
Nano Banana Pro is a deep learning model that can not only process and generate images, but also understand and respond to textual information found in prior images, thanks to its baked-in logic and intermediary prompting layers. The model can handle various tasks such as style transfer, object removal, text rendering, and code interpretation, while maintaining pixel-perfect text accuracy and embracing stylistic input. Its ability to handle character consistency across multiple reference images makes it a powerful tool for storytelling, brand consistency, and creating cohesive visuals. AI summary
Read more → -
Retro Diffusion's pixel art models, optimized for speed, high-quality, tile generation, and animation, are now available on Replicate, allowing developers to generate retro-style game assets, character sprites, and tilesets with various style presets and parameters. Four models are available: rd-fast, rd-plus, rd-tile, and rd-animation, which support different styles, input images, palette images, background removal, and seamless tiling. These models can be run from Python, JavaScript, and other languages using Replicate's SDKs. AI summary
Read more → -
Replicate is joining Cloudflare, expanding its resources and integrating with Cloudflare's Developer Platform to enhance its platform, while maintaining its existing API and model functionality. This move aims to leverage Cloudflare's existing infrastructure and expertise in building developer-centric products. The partnership enables Replicate to focus on building higher-level abstractions for AI, such as model orchestration and real-time model execution. AI summary
Read more → -
Datalab's Marker model extracts text from documents and images, including PDFs, DOCX, and images, with support for formatting tables, math, and code, and can pull specific fields using a JSON Schema. OCR detects text in 90 languages from images and documents, with high accuracy and fast processing times. Marker's performance was evaluated using the olmOCR-Bench benchmark, outperforming other OCR models, including GPT-4o, Deepseek OCR, and Mistral OCR. AI summary
Read more →