​
Extract insights from charts, tables, and images with Storytell’s Vision Ingest

Imagine trying to extract key insights from a thousand-page report, but half the critical data is locked inside images, charts, or infographics, rather than plain text. That’s where Storytell Vision Ingest steps in. This powerful capability within Storytell.ai is designed to instantly transform visual data into structured, actionable insights, ensuring that every file you upload—from slide decks to scanned documents—is fully understood, making its content truly query-ready. With Vision Ingest, you no longer risk skipping over crucial visuals or missing pieces of the puzzle. Storytell.ai ensures you get the complete picture.

​
What is Vision Ingest?

Storytell Vision Ingest is an advanced engine built into our platform, specifically designed for visual understanding. Its purpose is to analyze and interpret a wide array of visual elements embedded within your documents, such as charts, graphs, tables, images, and even flags. Unlike traditional Optical Character Recognition (OCR) which simply “reads” or extracts text characters from an image, Vision Ingest goes much further. It perceives and understands the content and context of your visuals. The result is a seamless integration of textual and visual information, processed with precision and delivered in a structured format that’s perfect for search, analysis, and AI-powered queries. The goal is to help users work with unstructured content more effectively by turning visuals into structured, query-ready data.

​
How Does Vision Ingest Work?

Storytell Vision Ingest operates using a sophisticated multi-modal pipeline, which includes a Vision-capable Large Language Model (LLM). This combination allows Storytell to not just see, but comprehend your visual data. Here’s a breakdown of the process:
1

Versatile File Upload

You can upload a variety of file types, including PDF, DOCX, PPTX, PNG, JPEG, or WEBP. Vision Ingest handles these mixed formats effortlessly, meaning you don’t need to reformat your files.
2

Smart Auto-Sharding for Large Files

To effectively process large documents, Vision Ingest automatically splits large files into manageable ‘shards’.
3

Visual Interpretation Beyond OCR

The Vision-capable LLM reads and understands your uploaded assets. This means it:
  • Describes and interprets elements like charts, tables, and images.
  • Can identify data trends in line or bar charts.
  • Can, for example, “understand the flags and find the countries from the flag images.”
4

Structured and Readable Output

The extracted visual information is transformed into markdown or HTML. This markdown-first output strategy is optimized for both human readers and AI querying, ensuring the output is clean, readable, and performs well with LLMs.
The intelligence of the Vision-capable LLM also extends to helping you formulate more effective questions. For example, when you upload an image, Storytell can analyze its content and suggest a more precise prompt to get the information you need, as shown below:
Think of Vision Ingest as an intelligent assistant that can look at the pictures in your documents and explain what they mean, just as it does with text.

​
Why is Vision Ingest so powerful?

Unlock Hidden Data

Many of the files our users upload include tables, maps, charts, or infographics that carry important information. Vision Ingest ensures this information is no longer skipped, providing a “more complete representation of the file.

Get Clearer Answers

By turning cluttered visuals to clear answers you can confidently ask questions like, “What’s the trend in this line graph?” or “Summarize the slide with the revenue chart.”

Achieve Faster Insights

Benefit from Smarter extraction, cleaner outputs, faster answers as visual data is automatically processed and made ready for querying.

Flexible File Handling

Upload various document and image types without worrying about pre-conversion.

​
Real-World Examples

Here are a few ways Vision Ingest helps you interact with your visual data:
  • Extracting Key Metrics from Dashboards and Reports: Business documents often contain dense visual information in dashboards or reports. Vision Ingest can process these, whether they are in slide decks, PDFs, or image files. For example, consider the following sales report dashboard: Example of a sales report dashboard with multiple KPIs and charts Instead of manually searching for data points, you can ask Storytell direct questions. Vision Ingest analyzes the visual elements to provide precise answers: If you ask: “What is the lead conversion ratio in Sales Report Dashboard Template?” Storytell provides: Similarly, if you inquire: “What is the lead to opportunity ratio?” Storytell responds:
  • Interpreting Infographics: Vision Ingest excels at understanding complex visuals like infographics. For instance, if you submit an image like the chart below, which displays languages and associated country flags: Example of a chart with language data and country flags You could then ask, “Based on this chart, identify the most spoken language and list the countries associated with it, as indicated by the flags.” Storytell, using Vision Ingest, processes the visual information to provide a direct answer: