This AI Tool Aids Enterprise Workflows by Analyzing Images

Generative AI startup Writer made waves in January after its Palmyra X V3 outpaced Google's PaLM 2. Now, it has unveiled a multimodal model capable of generating text based on images.

Palmyra-Vision is a multimodal model that can analyze images and generate content based on the objects and visuals in the images. It can extract handwritten text, classify objects and examine graphs and charts.

The multimodal model is designed to empower enterprise customers. The startup said that customers have routinely sought image-to-text capabilities to support their workflows.

Enterprises could use Palmyra-Vision, for example, to extract information from charts or help create Alt text descriptions. The model can even generate new content based on visuals like suggesting compliant copy for marketing images.

The model also has the ability to answer specific questions, so users could, for instance, ask if advertisements meet particular legal or regulatory requirements.

Credit: Writer

Palmyra-Vision was benchmarked against VQAv2, a dataset of open-ended questions on over 265,000 images that requires an understanding of vision, language, and common-sense knowledge. The model scored 84.4%, outperforming both OpenAI’s GPT-4V and Gemini 1.0 Ultra from Google.

Palmyra-Vision can be accessed via the Writer’s image analyzer app, which can be found in the startup’s library of prebuilt apps. The startup can also build custom apps to fit Palmyra-Vision to specific enterprise use cases.

Related:AI Startup Trumps Google in Stanford's Model Rankings

Writer’s multimodal unveiling follows recent updates to its text generation models to add support for multilingual capabilities in 30 languages including Spanish, French and Chinese along with English.

Last September, it raised $100 million in a Series B funding round that included participation by clients Accenture and Vanguard.