Turn any document into
fine-tuning data
Upload your document
Drop any PDF, DOCX, TXT, or Markdown file. We handle the rest.
Download your JSONL
AI generates Q&A pairs and formats them into OpenAI-ready JSONL.
Drag & drop a file
PDF, DOCX, TXT, MD
{"messages": [
{"role": "system", ...},
{"role": "user", ...},
{"role": "assistant", ...}
]}
{"messages": [
{"role": "system", ...},
{"role": "user", ...},
{"role": "assistant", ...}
]}
Upload your documents and get perfectly formatted JSONL training data for OpenAI fine-tuning. Ready in minutes.
The simplest way to fine-tune LLMs, for anyone
Three steps. No code. No data science degree required.
Fine-tuning is powerful. Preparing data shouldn't be painful.
OpenAI lets you fine-tune GPT on your own data — but getting documents into the right JSONL format is tedious and error-prone. We fix that.
The hard way
With JSONL for LLM
Supported formats
DOCX
.docx
TXT
.txt
Markdown
.md
OpenAI-ready output
Each line of your JSONL file contains a training conversation in the exact format OpenAI expects.
{"messages": [
{"role": "system", "content": "You are a helpful assistant that explains machine learning concepts."},
{"role": "user", "content": "What is gradient descent?"},
{"role": "assistant", "content": "Gradient descent is an optimization algorithm used to minimize the loss function in machine learning models. It works by iteratively adjusting parameters in the direction of steepest descent..."}
]}Frequently asked questions
What is JSONL fine-tuning data?
JSONL (JSON Lines) is the format required by OpenAI for fine-tuning models like GPT-4o and GPT-3.5. Each line contains a conversation with system, user, and assistant messages.
How are credits calculated?
Credits are based on the character count of your extracted document text. 1 credit per 5,000 characters, with a minimum of 1 credit per document.
What file types are supported?
We currently support PDF, DOCX (Microsoft Word), plain text (.txt), and Markdown (.md) files up to 10MB.
How long does processing take?
Most documents are processed within 1-3 minutes, depending on size. You can track progress in real-time on the document detail page.
What model generates the training pairs?
We use GPT-4o-mini via OpenRouter to generate high-quality question-answer pairs from your document content.
How do I use the JSONL file with OpenAI?
Go to platform.openai.com → Fine-tuning → Create. Upload your downloaded JSONL file, select a base model (like gpt-4o-mini), and click Create. OpenAI handles the rest — training usually takes 10-30 minutes.