News · January 5, 2021
OpenAI shows DALL·E, a GPT-3 variant that draws images from text
OpenAI has shown a model that turns short text descriptions into images, alongside a companion model for classifying images. It is research for now, but it hints at where creative and visual tooling could go.
What happened
- OpenAI introduced DALL·E on January 5, 2021, describing it as a 12-billion-parameter version of GPT-3 trained on text and image pairs to generate images from captions, according to its research post.
- OpenAI said it used a second model, CLIP, to rank the best of 512 generated samples for each caption shown in its examples.
- Axios reported that CLIP can classify images against any set of visual categories while needing less training and computing power than existing methods.
- Axios also noted that the models are imperfect and that DALL·E's results depend heavily on how a prompt is worded.
Why it matters
This is a research release, not a product, so there is nothing to integrate yet. What matters for product teams is the direction: the same transformer approach behind GPT-3 now handles images, and prompt wording becomes a design skill in its own right.
For marketing teams that produce many ad and store creatives, it is worth keeping an eye on this space and on CLIP-style classification, which could make tagging and searching large creative libraries cheaper. For now we would not change budgets or workflows because of it.
Sources
This post is CMM’s own summary and commentary on publicly reported news. Facts are attributed to the sources linked above; please read them for full details. CMM is not affiliated with these publishers or with the companies mentioned, and trademarks belong to their owners. Prepared with AI assistance. Corrections or requests: [email protected]. See our editorial policy.
Want help with this?
Tell us what you need, and we will suggest the right team model on a short call.
Talk to us