A return to hand-written notes by learning to read and write

Use cases & UX for handwriting capture

• Many see value for teachers and presenters: write quickly on a board or tablet, have the system “clean up” handwriting while preserving a handwritten look.
• Others suggest skipping handwriting entirely: use keyboards, projectors, or large touchscreens with typed text, but critics say this disrupts flow, eye contact, and fast sketching.
• Several mention current tools (iPad Notes, note‑taking tablets, OCR apps) that already neaten handwriting or convert it to text with decent accuracy.
• Some prefer analog workflows (paper, whiteboards, fridge whiteboards) plus occasional photo/OCR as a low-friction compromise.

Handwriting vs digital fonts

• Debate over replacing messy handwriting with perfect fonts: proponents value uniformity and legibility; opponents stress loss of personality, flexibility for arrows/diagrams, and subject‑specific letter tweaks.
• Some view cleaned-up handwriting that still “looks like you” as an ideal middle ground.

Improving handwriting & tools

• Several argue that simply practicing, slowing down, and using block or non-joined letters significantly improves legibility.
• Others recommend fountain pens, gel pens, or specific grips to force slower, more intentional strokes; some report big gains, others say tools don’t overcome dysgraphia or poor motor skills.
• Resources mentioned include calligraphy/italic manuals, handwriting repair approaches, and special practice sheets.

OCR and technical quality

• Tesseract is praised for book scans and invisible OCR in PDFs, but criticized for poor performance on screenshots and non-English scripts.
• Users are impressed by modern phone/iOS handwriting recognition and ChatGPT OCR, though accuracy remains around 90–95% and needs proofreading.
• Some want open-source, offline handwriting OCR that can convert notes to markdown reliably.

Privacy, openness, and data

• A few are skeptical that the project is a way to harvest handwriting data for training.
• Others counter that the model and code are open, runnable offline, and there is no built‑in data collection; they frame it as typical research, not a product.

Risks, applications & broader reflections

• Concerns about enabling forged signatures or fake handwritten manuscripts; others note the model isn’t generative but acknowledge related work exists.
• Potential benefits suggested for education, remote teaching, preserving old documents, and historical handwriting transcription.
• Several discuss the decline of everyday handwriting due to computers/phones, yet still find cognitive value in handwritten note‑taking and whiteboard work.