In our visually rich digital world, Images convey vast amounts of information, tell stories, and evoke emotion. But what happens when these images are invisible to a segment of your audience? For millions of people who are blind, have low vision, or use assistive technologies like screen readers, the internet can become a landscape of missing pieces without a simple yet powerful feature: alternative text, or “alt text.“

Alt text is a concise, written description of an image that screen readers announce aloud, providing the essential context and meaning that sighted users perceive visually. It’s not just an accessibility feature; it’s a cornerstone of an inclusive web, ensuring everyone can engage with online content equally. It also plays a role in search engine optimization and helps when images fail to load. Yet, crafting high-quality, accurate alt text for every image, especially across numerous web pages or large documents, can be a significant and often overlooked challenge for content creators.

To help bridge this gap and streamline this vital process, we are pleased to announce the launch of a new solution: an AI-Powered Alt Text Generator. This application, currently in an experimental phase and available free of charge, is designed to assist users in creating descriptive alt text for images more efficiently, empowering you to make your content more accessible with greater ease.

Meet the AI Alt Text Generator: Your Accessibility Assistant

Our Alt Text Generator utilizes advanced artificial intelligence – specifically, Google’s Gemini API (an AI model capable of image comprehension and descriptive text generation) – to provide a strong starting point for your alt text.

Key features include:

• Efficient Image Handling: Upload up to 50 images (JPEG, PNG, GIF, WEBP) at once. This batch processing capability is particularly beneficial when working with numerous photos for reports, website updates, blog posts, or when you need to retrospectively add alt text to previously published visual content.
• Multilingual Support: Select the desired target language for the generated alt text. (19 languages available)
• Batch Generation: Process all uploaded images efficiently with a single command. (The tool takes tiny breaks to process all the images!)
• Confidence Scores: The tool assigns confidence scores (High, Moderate, Needs Review) to each AI-generated description, aiding in the prioritization of human review.
• Direct Editing: Review and modify AI suggestions directly within the tool’s interface.
• Simple Export: Download finalized alt text in a .txt file format.
• Copy Alt Text: Alt text for each photo can also be copied and pasted into your tool where you are working with images.

How To Use Alt Text Generator

1. When you first visit the AI Alt Text Generator, you’ll see a clean interface ready for action. The page explains its purpose: to improve web accessibility by generating descriptive alt text for your images, with the capability to upload up to 50 images at a time. Key controls like ‘Choose Files,’ ‘Target Language for Alt Text’ (defaulting to English), and generation/download buttons are visible. Initially, a message indicates ‘No images uploaded yet,’ prompting you to begin.

2. Once you’ve selected your images using the ‘Choose Files’ button, they appear as individual cards. This screenshot shows 18 out of a possible 50 images loaded. Each card displays the filename (e.g., ‘import_photos_2801.jpg,’ ‘natural-bridge.jpg’), file size, and the currently set language for generation (English, in this case). The text area for ‘Generated Alt Text’ is empty, inviting you to either generate text for individual images or use the main ‘Generate Alt Text (18)’ button at the top to process all of them.

3. Our tool is designed for global inclusivity. Before generating, you can easily select your desired language for the alt text from the ‘Target Language for Alt Text’ dropdown menu. This screenshot shows the dropdown expanded, with English selected, and a list of other available languages including Spanish, French, German, Japanese, Chinese (Simplified), Italian, Portuguese, Korean, Hindi, and Arabic. The ‘Generate Alt Text (18)’ button indicates images are ready once the language is set.

4. After clicking ‘Generate Alt Text,’ the AI gets to work! This screenshot demonstrates the results. The same two images (‘import_photos_2801.jpg’ and ‘natural-bridge.jpg’) now have descriptive alt text filled into their respective editable text areas. For example, the band photo’s alt text reads, ‘A black and white photo shows four men in smart casual attire…’ and the natural bridge description is ‘A paved walkway lined with stone walls…’ Character counts are displayed, and each card now features a ‘Regenerate (English)’ button if you want the AI to try again. You can also see options to ‘Show Confidence’ scores and filter your images. It is highly encouraged to add context and other detail to this “draft” that the AI may have missed and/or incorrectly depicted (cultural nuances for example).

Further details on utilizing the tool and understanding its features are available on our Guidance & Confidence Scoring page.

The “Vibe Coded” Story: An Experiment in AI-Driven Development

Beyond its function in generating alt text, this tool represents a significant experiment for us, enabling us to take a significant step in the realm of Artificial Intelligence. As detailed on our “About” page, the application was developed using a method termed “vibe coding.“ This involved guiding AI with natural language prompts and a general vision, using the Lovable.dev platform, as an alternative to traditional line-by-line coding.

Given its experimental nature and reliance on the Gemini API (for which we are currently covering the operational costs), we aim to gather data on the real-world expenses of AI-driven alt text generation and application development in general, and share with the community. Your utilization of the tool will contribute valuable insights to this research.

The Power of AI, The Necessity of Human Oversight: A Partnership for Effective Alt Text

A critical aspect of this tool is understanding its intended role. While artificial intelligence offers significant potential to expedite laborious tasks, this generator is designed to augment human judgment, not supplant it. It functions as an efficient assistant, providing initial drafts. The user’s expertise, cultural understanding, and contextual knowledge remain essential for refining and approving the alt text to ensure it is both meaningful and accurate.

As our on-site guidance on our Tips & Considerations page further elaborates:

Benefits of AI Assistance

• Streamlined Workflow & Reduced Effort: Substantially decreases manual labor and cognitive load, particularly for large image sets or for users who find extensive typing challenging.
• Enhanced Quality Control: Confidence scores assist in prioritizing review efforts on descriptions where the AI indicates lower certainty.
• Democratizing Access: Offers AI-powered tools for multiple languages, including many from the Global South, potentially fostering greater digital inclusion.
• Valuable Learning Insights: Observing the AI’s performance can deepen understanding of effective alt text characteristics.

Key Considerations: The Indispensable Role of Human Review

• Risk of Overreliance: Even AI suggestions with “high confidence” can overlook vital context or nuance. Critical human review is always necessary.
• Accuracy in Diverse Languages: AI models may exhibit lower accuracy or cultural sensitivity for languages underrepresented in their training data. Native understanding and cultural awareness are paramount during review.
• Potential for Bias: AI systems can reflect biases present in their training datasets. Vigilant human oversight is required to ensure fair, equitable, and accurate descriptions.
• Legal Compliance: Human review and approval are crucial for meeting accessibility standards and requirements.

It is important to remember: the AI generates; the user validates and perfects. Your role encompasses ensuring accuracy, contextual relevance, cultural sensitivity, conciseness, and the avoidance of redundancy.

Tested and Validated: Insights from Juanita Lillie

We were privileged to have Juanita Lillie test the new AI-Powered Alt Text Generator using VoiceOver on her iPhone.

Juanita, based in Michigan, is a lifelong blindness advocate with years of experience collaborating with others to enhance interdependence and access. We first connected with Juanita several years ago through her impactful advocacy in the assistive technology space. While now employed full-time, she continues to lead and support various advocacy projects and manages A Blind Advocate,  a platform where she shares real-life experiences, resources, and straightforward insights.

Juanita provided detailed and thoughtful feedback, noting her appreciation for the tool. She observed that while tools exist for blind and low-vision individuals to create alternative text, this tool is notable for enabling sighted individuals to easily generate alt text as well. She particularly valued its availability for general use.

We are very grateful for Juanita’s time, perspective, and continued collaboration. Her positive feedback affirms the tool’s potential utility and user-friendliness.

We Invite Your Participation and Feedback

We invite you to explore the AI Alt Text Generator.

As previously stated, the tool is free to use during this experimental phase. Your usage and, importantly, your feedback are highly valuable. We welcome your input on:

• The tool’s ease of use.
• The quality of the AI-generated alt text.
• Any issues encountered, particularly with assistive technologies.
• Suggestions for improvement.

Kindly share your feedback via our Contact page.

We are pleased to offer this tool to our community. It is our belief that by combining the capabilities of AI with essential human oversight, significant progress can be made in enhancing digital accessibility for all users. We anticipate your feedback.

---