From uploading a handwritten page to getting your account set up and running, here's everything you need to know — including how to fix the installation errors people run into most.
Three steps take a handwritten document from raw image to clean, editable text.
Upload a JPG, PNG of your handwritten or historical document. You can upload a single page or a multi-page file, and low-quality or faded scans are supported.
Select the OCR model trained for your document's language and era — German handwriting, French manuscripts, or a general multilingual historical model.
The AI extracts the handwritten text with up to 90% accuracy. Review the result side-by-side with your original scan, then export it.
Setting up TranscribeCR takes a few minutes. Here's what happens at each step.
Download and install TranscribeCR, then create a free account with your email address. This account is where your uploaded documents, extracted text and credit balance are stored.
Open the app and sign in with the account you just created. You'll stay signed in on this device, so this is a one-time step unless you sign out.
The very first time you sign in, TranscribeCR downloads and initializes its AI OCR models locally. This can take a few minutes depending on your internet speed — this is expected and only happens once. Keep the app open and connected until it finishes.
Every new account starts with 5 free credits, each good for one document conversion. Your remaining balance is shown in the app header, and you can upgrade any time once you've used them.
The most frequent issues people run into during install or first sign-in, and how to resolve them.
This error occurs when PyTorch cannot load properly due to version conflicts, missing system dependencies, or architecture mismatches. The same code may work on one system but fail on another due to environmental differences.
Common Causes:
Step-by-Step Fix:
python --version (Must be 3.8-3.11, preferably 3.10)python -m pip install --upgrade pippython -c "import platform; print(platform.architecture())" (Must be 64-bit)nvidia-smi to see CUDA versionpip uninstall torch torchvision torchaudio torchtext -y (run twice)For CPU-Only Systems (Recommended for most users):
pip install torch==2.2.2 torchvision==0.17.2 torchaudio==2.2.2 --index-url https://download.pytorch.org/whl/cpuFor GPU Systems (CUDA 11.8):
pip install torch==2.2.2 torchvision==0.17.2 torchaudio==2.2.2 --index-url https://download.pytorch.org/whl/cu118For GPU Systems (CUDA 12.1):
pip install torch==2.2.2 torchvision==0.17.2 torchaudio==2.2.2 --index-url https://download.pytorch.org/whl/cu121Verification: Create a test script to confirm installation:
python -c "import torch; print(f'PyTorch {torch.__version__}'); print(f'CUDA available: {torch.cuda.is_available()}')"If still facing issues:
This is almost always a missing system dependency (like a runtime library) rather than a problem with the installer itself.
The first launch downloads the OCR models, which normally takes a few minutes. If it's been stuck for much longer than that:
This happens when your GPU doesn't have enough memory to process the image. Common fixes:
--cpu flag or set environment variable)This usually indicates a model mismatch or preprocessing issue:
PyTorch is not installed or not found in your Python environment:
pip list | grep torch to check if PyTorch is installedsource venv/bin/activate (Linux/Mac) or venv\Scripts\activate (Windows)This is almost always a missing system dependency (like a runtime library) rather than a problem with the installer itself.
The first launch downloads the OCR models, which normally takes a few minutes. If it's been stuck for much longer than that:
This is usually a cached session or an unverified email address.
New accounts should start with 5 free credits automatically.
The installer needs permission to write files to your system folders.
This is a document issue rather than an installation issue, but it's asked often enough to include here.
Create your account, sign in, and start converting handwritten documents with your 5 free credits.