I need help with study materials
retained text
I've been using AI a lot recently to study. I've taken PDFs and made summaries with AI. I'm having problems with the size of the files. I realized that most harnesses do not support large files. For example, for PDF files, the maximum size is usually 5 megabytes, while for video files it is 50 MB. Since my files are larger than this, I have processed them on my computer, extracting the text and summarizing it from there. But I realized that the quality of files processed on my computer is much lower than when processed natively by AI. The PDFs are okay, mainly using MIMO 2.6 Pro and DeepSeek Harness, but the videos are horrible. When I process videos on my computer, the result is terrible — with many slides, the AI can't convert them into text. I recently discovered that Google's AI Studio can process large files, both video and PDF. But it has its limitations — I have to upload one file at a time. In some subjects, I have more than 100 classes and dozens of PDFs. I can't do this kind of processing, even using Antigravity, which is from Google and has the same size limitation. I thought about creating an MCP to use with Google AI Studio, but I saw that it had the same file size limitation. I looked for other AIs to see if I could do it natively. I found out that you can use MIMO Desktop to do this processing, but when I tried to install it, it said the version doesn't match my region. Luckily, I didn't subscribe before trying. The problem is that the material is in Portuguese, and I don't know how good the text extraction and processing would be in another AI when it's not in English. By the way, I also don't have a lot of money. Does anyone have any suggestions on how I can solve this problem? I'm using Google AI Pro for Students and OpenCode Go. Update 1. I took one of the PDFs that had 222 pages, transformed each page into an image, and used DeepSeek 4.1 to extract the text, and I got a better result for the part I was having problems with (images). And the result was very good. Unfortunately, DeepSeek is not very good with Portuguese, and it was not the same as when I used MIMO 2.6 Pro or Gemini 3.8 flash (they kept the same words that were in the PDF). The problem is that the process took 3 hours, used more than 15 million tokens, and cost about 80 cents. I could have done all the PDFs that way, but unfortunately the result was not good (for Portuguese). Using MIMO 2.6 Pro would have drained my Open Code Go account. I might try with Mimo 2.6 flash later.
document fetched 2026-10-25 00:18 · canonical https://reddit.com/r/AI_Agents/comments/1xeglgu/i_need_help_with_study_materials