OpenSea has added alignDRAW, the historic text-to-image project led by AI researcher Elman Mansimov, to its Flagship Collection. The selection places a 2015 generative-AI experiment inside OpenSea’s curated NFT reserve nearly a decade after the underlying images were created. OpenSea announced the addition on September 17, describing alignDRAW as the first text-to-image AI model and highlighting its distinctive 32-by-32-pixel outputs.
The project’s OpenSea collection page lists 2,709 works generated from prompts used with Mansimov’s model. The images themselves date to 2015, while their NFT release came years later through Fellowship in November 2023. That distinction makes the Flagship inclusion a curation event for an existing collection rather than a new mint or the launch of a newly developed AI system.
A 2015 Research Project Finds an On-Chain Archive
Mansimov developed alignDRAW with Emilio Parisotto, Jimmy Lei Ba and Ruslan Salakhutdinov, with the research published in the 2015 paper “Generating Images from Captions with Attention.” The model generated images iteratively from natural-language descriptions by drawing patches on a canvas while attending to relevant words in the prompt. Its outputs were primitive by current generative-AI standards, but they demonstrated that a neural network could construct novel scenes from written descriptions rather than simply retrieve matching images.
The surviving collection consists of 2,709 images generated from 36 unique text prompts, with each output measuring only 32 by 32 pixels. Those technical limitations are now part of the collection’s historical identity rather than something being concealed by later enhancement. Fellowship preserved the original pixelated files when bringing the works on-chain, and the collection subsequently became fully on-chain.
OpenSea has previously highlighted Mansimov’s work as an early milestone in text-to-image generation, while its latest announcement goes further by calling alignDRAW the first text-to-image AI model. That historical claim is OpenSea’s characterization and should be distinguished from the narrower claim made in the original academic paper, which documented a model that generated images from natural-language descriptions without itself establishing a comprehensive priority record for the entire field.
OpenSea Expands Its Digital-Art Reserve
OpenSea launched its Flagship Collection in September 2025 as an NFT reserve backed by a commitment of more than $1 million for acquisitions. The marketplace says a committee of employees, advised by external specialists, selects works based on their perceived long-term relevance to digital art and NFT culture. Adding alignDRAW therefore represents a curatorial decision by OpenSea rather than a new technical capability or an endorsement of the collection as an investment.
The selection also illustrates how early AI research outputs have acquired a second life as digital art objects. What began as an academic experiment was later exhibited at Paris Photo and around Art Basel before becoming a fully on-chain Ethereum collection. Its significance within OpenSea’s Flagship program rests largely on provenance and technological history, not on the visual fidelity expected from contemporary image-generation systems.
The next concrete development to watch is how OpenSea documents the specific alignDRAW work or works held in its Flagship reserve and how the marketplace continues building the collection around historically significant digital art. For alignDRAW, the September addition strengthens its curation history without changing the underlying 2015 work, its supply or the mechanics of the existing NFT collection.