These features need a language, speech or translation model, which StudyGrasp does not use today. We would rather tell you what stands in the way than fake them. Tell us which ones you want and we will follow the votes. StudyGrasp stays free.
Voting needs a free StudyGrasp account. You will be asked to log in or sign up first.
Coming soon
Essay scoring and written feedback
A score estimate and personal written comments on an essay.
What works today. Rubric check shows which rubric criteria your draft has evidence for. It is a checklist, not a grade.
What is in the way. Judging whether an argument is good needs a language model. A rule-based score looks authoritative but would mislead, so only the checklist ships.
The possible fix. A free, private model good enough to grade (today: a multi-gigabyte in-browser model with weak quality, or a paid hosted API that sends drafts to a third party). Revisit if a small open model becomes good enough and the owner relaxes the no-language-model rule.
Coming soon
Audio and video files to study sets
Upload an MP3 or MP4, or paste a YouTube link, and get a transcript automatically.
What works today. Lecture to notes turns a transcript or caption file (.srt, .vtt) into a summary, key terms and flashcard ideas. Dictate live under My content > Record a lecture.
What is in the way. Speech-to-text models are 40 to 80 MB or more, slow on long lectures in a browser and weaker on technical terms; hosting one costs money and means uploading audio. YouTube has no free official caption API, and scraping breaks the platform's terms.
The possible fix. An optional, lazy-loaded 'Transcribe on this device' button using a quantised Whisper model (MIT), showing the download size first, keeping audio on the device. Needs a design pass for large files and tests on low-end devices.
Coming soon
Word-problem solver
Type a problem in plain English and have it set up and solved.
What works today. Step solver works arithmetic, equations, percent, interest, loans, averages and unit conversions, one step at a time.
What is in the way. Turning English into an equation needs language understanding. The rule-based solver only handles fixed problem types and says so rather than guessing.
The possible fix. More problem families added by rules (systems of equations, derivatives, geometry, stoichiometry), or a model for free text. A CAS library (Nerdamer or Algebrite, MIT) could widen coverage at some bundle cost.
Coming soon
Open-ended tutor conversations
Ask anything in your own words and get reasoned, conversational explanations.
What works today. Tutor chat answers from your course material and your own uploads and shows where each answer came from.
What is in the way. Reasoned conversation beyond the source text is what a language model does. Hosted models cost money and see student questions; in-browser models are 1 to 4 GB with noticeable errors, and some licences restrict use.
The possible fix. An owner decision to change the no-language-model rule, with funding or donated hosting and a safety review. Meanwhile: better retrieval and follow-up handling in the rule-based tutor.
Coming soon
Paraphraser and sentence rewriting
Rewrite whole sentences in new words.
What works today. The Writing coach flags grammar slips, misspellings (with a real spell checker) and wordy phrases, with a plainer option you apply yourself.
What is in the way. Rule-based rewriting changes meaning too often, and automatic rewriting invites copying, which conflicts with the Writing coach's academic-integrity stance.
The possible fix. A language model with guardrails, or leave it parked. Better rule coverage (more phrase rules, other-language dictionaries) is incremental.
Coming soon
Study in 20+ languages
Menus, notes, flashcards and chat in many languages.
What works today. Menus and common buttons in Spanish, French and Portuguese (beta). In Chrome on a computer, Lecture to notes can translate a summary on your device.
What is in the way. Translations need native review (French and Portuguese are still unreviewed), course packs are in English, and good machine translation of content needs a translation model. Open models are 20 to 60 MB per language pair and non-commercial licences (NLLB) do not fit.
The possible fix. Native reviewers per language (one catalogue object each), optional on-device translation through the browser Translator API in more places, optional per-language model downloads later.
Coming soon
Fully generated explainer videos
Narrated videos written and animated automatically.
What works today. Explainers and Audio recaps build narrated slides and audio from your course, and Lecture to notes summarises a transcript.
What is in the way. Video generation needs media models and heavy compute that neither run in a browser nor are free to host. A consistent narrator voice would need an open text-to-speech engine with its own licences.
The possible fix. Nothing realistic within the free and private constraints. The slide-and-narration Explainers are the practical version; an open voice such as Piper could make narration consistent.
Everything else already works: cited tutor answers, flashcards, quizzes, practice tests, your own uploads, photo to notes, lecture notes, the step solver, the writing coach and more. See What's new.
