April 2026 was packed with improvements across the board. We introduced new ways to control your flashcard experience, added an AI model selector for slides and video lectures, brought two new AI models to the platform, redesigned the mobile experience, upgraded to better AI voices, and shipped dozens of bug fixes to make everything more reliable. Here's the full rundown.

---

Major Features

Flashcard Density Options

You now have full control over how many flashcards are generated from your study material:

• Low (~30 cards) — perfect for quick reviews of the most important concepts.
• Medium (~60 cards) — a solid balance between coverage and study time.
• High (~120 cards) — deep coverage for thorough exam prep.
• Very High (400+ cards) — exhaustive, leave-nothing-uncovered study sessions for when you need to know everything.

Choose the density that fits your study style and how much time you have before the exam.

Refine Deck with AI

Already have a flashcard deck but want it tweaked? Now you can tell AI to reshape your entire deck:

• Make cards harder, simpler, more detailed, or change the focus entirely.
• Quick suggestions like "Make it harder" and "Simplify the language" for one-click refinement.
• Available after studying 5 cards or from the editor tab.
• Large decks are processed in batches for reliable results.
• Premium feature with 3 refinements per deck.

AI Model Selector for Slides & Video Lectures

You can now choose which AI model generates your slides and video lectures. Pick from Grok 4.3, GPT 5.4 Mini, or Gemini 3 Flash — each produces different results, so experiment to find your favorite.

System prompts automatically adapt to whichever model you pick, so generations come out polished regardless of your choice.

Customizable Add Content Modal

The Add Content modal now lets you customize AI Video Lectures, Podcasts, and AI Slides before generating — just like flashcards. Pick a preset, length, voice, custom instructions, and more, all from one place.

Recent Chats Strip

Chats on pages, PDFs, videos, podcasts, and recordings now show your 5 most recent conversations for that piece of content right at the top. Jump back into a previous conversation with one click, or hit "Show more" to see the full history.

---

New AI Models

GPT 5.5

OpenAI's GPT 5.5 is now the recommended premium OpenAI chat model. It delivers deeper reasoning, better coding help, stronger research, and more structured long-form answers. GPT 5.4 Mini remains as the free default for everyday chats.

Kimi K2.6

Moonshot AI's Kimi K2.6 is available as a premium model in chat, AI Slides, and AI Video Lectures. It's a 1-trillion-parameter open model with a massive 262k context window, great for working with large documents. Routed through Cloudflare for fast US-based inference.

Grok 4.3

Grok 4.3 from xAI is available across chat and AI Slides as a thinking model, with lower pricing for faster, more affordable generations. Grok now uses xAI's newer Responses API, improving tool calling reliability and support for reasoning traces.

---

Faster Flashcard Generation

Study mode questions (true/false, multiple choice, short answer, fill-in-the-blank) are now generated alongside your flashcards in a single step — no more waiting for a second processing pass. Questions also test different angles instead of repeating the same fact four ways.

---

Voice & Audio Upgrades

New AI Voices for Podcasts

Podcast and audio narration now uses Google's Gemini 3.1 Flash TTS — noticeably more natural, expressive voices with better pacing across 70+ languages.

Upgraded Voice Assistant

The voice assistant now uses xAI's Grok Voice Think Fast model — faster responses, more natural turn-taking, better handling of accents and background noise, and more reliable tool use when you ask it to pull information from your PDFs, videos, or the web.

---

Image Generation Upgrade

AI image generation across the platform now uses OpenAI's GPT Image 2. You'll notice higher quality illustrations, diagrams, and cover art in flashcards, chat, podcasts, AI Slides, video lectures, and pages. Editing images with follow-up prompts now supports combining up to 16 reference images in one request.

---

AI Video Lecture Improvements

Video Length Selector

Choose how long your AI Video Lecture should be:

• Short (1–2 min) — quick concept overview.
• Medium (3–5 min) — balanced explanation with examples.
• Long (8–12 min) — in-depth deep dive.

Better Research

Video Lectures now research your source material more thoroughly before generating, resulting in better explanations and more relevant examples.

---

AI Slides Quality Boost

AI Slides now automatically renders every slide as an image and visually checks each one for overlaps, text cutoffs, and layout issues before finalizing. The result: higher quality, more reliable slide decks.

---

Mobile Redesign

We overhauled mobile responsiveness across the entire app:

• The page editor now uses full screen width on mobile instead of being cramped.
• Login, register, and settings pages display properly on small screens.
• The Explore content grid shows a single column on mobile instead of tiny, unreadable columns.
• Header buttons collapse to icon-only on small screens to save space.
• Landing pages, feature pages, and marketing pages scale properly on all devices.

---

Chat Improvements

Research from Chat

You can now ask Chat to start a deep research session — just like asking it to make a podcast, video lecture, or slide deck. The research runs in the background and shows up inline in chat when it's ready. Chat also asks clearer follow-up questions before starting, so the final report better matches what you want.

Better LaTeX Rendering

Fixed LaTeX occasionally rendering as raw source code when an equation spanned multiple lines or contained matrix environments. Multi-line math now displays correctly.

Shared Chat Images

Images uploaded into chat now render as inline thumbnails everywhere — both in your own chats and in shared chat links — instead of showing up as a plain file name.

---

Upgrade & Trial Experience

3-Day Free Trial

Yearly plan subscribers can now try Ultimate free for 3 days before being charged. The pricing page clearly highlights this so you know before you click.

Smarter Upgrade Prompts

Upgrade prompts now appear at the right moment — after completing a study session, finishing an exam, or hitting a limit — not randomly during your workflow. Removed passive upgrade popups that could appear mid-session while studying.

Pre-Cancel Features Discovery

When you tap Cancel Subscription, we now show a quick rundown of premium features you haven't tried yet — AI Video Lectures, Podcasts, Recordings, Deep Research, and more — with "Try it" buttons so you can give them a spin before deciding.

Trial Ending Notification

Added an in-app heads-up when your free trial is about to end, so you have a clear chance to keep Ultimate before losing access.

---

Task Retry & Error Messages

• You can now retry failed flashcard and podcast generation directly from the task dropdown — no need to start over.
• All generation tasks (flashcards, podcasts, slides, video lectures, recordings) now show clear, actionable error messages instead of generic failures.
• Background tasks no longer get stuck in a "Processing" state when something times out.
• Added a feedback prompt after study sessions, exams, and flashcard decks so you can tell us how it went.

---

New Help Guides

We added and updated several help guides this month:

• Video Lecture Length — picking between Short, Medium, Long, and when each is best.
• Choosing an AI Model — when to pick GPT, Gemini, Grok, or Claude.
• Refine Flashcards with AI — reshaping an entire deck in one action.
• Add Content — uploading files and creating content with full customization.
• Recent Chats — the recent chats strip and jumping between conversations.
• Hard Questions — flagging difficult flashcards and how cards auto-resolve.
• Commercial Use — what you can do with content you create on Scholarly.
• Updated guides for AI Video Lectures, AI Slides, Podcasts, Flashcards, AI Chat, and Getting Started.

---

Bug Fixes

• Fixed quiz questions not appearing when switching to the Questions tab.
• Fixed flashcard study view getting stuck on a loading spinner after an error.
• Fixed the "Edit Flashcards" button opening an empty modal instead of the refine prompt.
• Fixed AI Video Lectures and AI Slides failing when creating from more than 3 uploaded PDFs.
• Fixed prompt-only AI Slides and Video Lecture requests failing before generation could begin.
• Fixed failed flashcard creation requests counting against your daily upload limit — failed uploads are now refunded.
• Fixed fill-in-the-blank questions appearing with the blank in awkward positions.
• Fixed exams not always including the number of questions you asked for.
• Fixed certain PDFs failing when one page was broken — Scholarly now skips the bad page.
• Fixed the video completion event firing multiple times when a video ends.
• Fixed website-to-PDF imports failing on sites like Medium and StackOverflow — we now automatically retry through our rendering fallback.
• Fixed duplicate share records causing visibility toggles to sometimes not work.
• Fixed keyboard shortcuts for flashcards still working behind the upgrade modal.
• Fixed completed exams not always showing up in study totals right away.
• Fixed AI chat PDF requests quietly timing out instead of showing a clear error.
• Fixed a login redirect loop for users with expired sessions.
• Improved PDF processing reliability with better retry and fallback handling.
• Fixed podcast generation failing when the text-to-speech provider flagged benign passages — we now retry with adjusted phrasing.
• Fixed research sessions showing a false "Couldn't connect to the server" error during brief network blips.
• Reduced noise from harmless network interruptions so real errors surface faster.
• Fixed a rare crash when previewing very long documents or transcripts.
• Fixed the Filter dropdown on the All Content page not opening when clicked.
• Fixed audio and video players failing silently when a media link expired — they now show an error and offer to retry.
• Fixed an editor crash when using word-level cursor shortcuts near the start of a page.
• Various stability and performance improvements across the platform.

---

Looking Forward

April 2026 was all about giving you more control — over flashcard density, AI model choices, video length, and how you customize content before generating it. Combined with three new AI models, upgraded voices, better image generation, and a completely revamped mobile experience, Scholarly is more flexible and reliable than ever.

If you have any questions or feedback, reach out to us at [email protected].