Generate PDF from template in Python: Jinja2 with WeasyPrint, or a template API with requests

The two Python routes: Jinja2 with WeasyPrint, or an HTTP call

To generate a PDF from a template in Python, fill an HTML template with Jinja2 and convert the result with WeasyPrint; both are open source and the whole job is about five lines. The alternative is to keep the template in a hosted editor and send JSON to a template API with the requests library.

The first route is enough for most Python projects where a developer owns the layout. The second is for teams where someone else does. If you are still choosing between approaches, the overview of the three template routes comes first.

Route 1: Jinja2 and WeasyPrint, step by step

Prerequisites. WeasyPrint's installation page lists Python 3.10 or later and Pango 1.44 or later for version 70.0. Pango is a system library, not a pip package: on Linux it comes from your package manager, and on Windows the documentation installs it through MSYS2. Check this before you plan a deployment.

1. Install the renderer: pip install weasyprint. Install Jinja2 from PyPI as well.
2. Create the template environment: env = Environment(loader=FileSystemLoader("templates"), autoescape=select_autoescape())
3. Load the template file: template = env.get_template("invoice.html")
4. Merge the data: html = template.render(invoice_number="INV38379", total_amount=82542.56)
5. Render the PDF: HTML(string=html).write_pdf("invoice.pdf")

The imports are from jinja2 import Environment, FileSystemLoader, select_autoescape and from weasyprint import HTML. Called with no arguments, write_pdf() returns the PDF as bytes instead of writing a file, which is what you want in a web view. Page size and margins are plain CSS: the documentation's example is CSS(string='@page { size: A3; margin: 1cm }'), passed through the stylesheets argument of write_pdf.

Route 2: a template API over HTTP with requests

Prerequisites. A CraftMyPDF account, a template designed in its editor, the template ID and your API key. The Free plan covers 50 PDFs a month and 3 templates (pricing page, 3 October 2026).

1. Build the payload: payload = {"template_id": "05f77b2b18ad809a", "export_type": "json", "data": data} where data is your dictionary. The API reference accepts a JSON object or a stringified one, and its own sample uses json.dumps(data).
2. Send it: response = requests.post("https://api.craftmypdf.com/v1/create", headers={"X-API-KEY": api_key}, json=payload)
3. Read the link: pdf_url = response.json()["file"]. The response also carries status and transaction_ref.
4. Fetch the PDF from that URL before it expires, or set "export_type": "file" to receive the PDF bytes directly in response.content.

Optional parameters from the reference include expiration (minutes the link stays valid, default 5), output_file (file name, default output.pdf) and version (a specific template version). The CraftMyPDF API guide covers the asynchronous endpoint and regional hosts.

Worked example: one invoice on both routes

Take the invoice data from CraftMyPDF's reference sample: invoice number INV38379, date 2021-09-30, currency USD, total 82542.56.

On route 1, invoice.html contains placeholders such as {{ invoice_number }} and {{ total_amount }}. template.render(...) returns the filled HTML string and WeasyPrint writes invoice.pdf to disk. Nothing leaves your server.

On route 2, the same four values go in the data dictionary. The service returns a JSON body whose file value is a link to the PDF. One call uses one credit, whatever the page count.

A mistake to avoid on each route

• WeasyPrint: passing the HTML as a positional argument. HTML('<h1>foo') is read as a file name. A string in memory must be passed by name, as HTML(string=...). The documentation calls this out directly.
• WeasyPrint: rendering untrusted HTML. The documentation warns that untrusted HTML or CSS can cause security problems. Keep autoescape on in Jinja2 for user-supplied values.
• Template API: saving the link instead of the file. The default link lifetime is 5 minutes. Download the PDF or raise expiration.
• Either route: checking only short documents. Test a long table before release; see how page breaks go wrong.

If generation fails in production, the failed generations checklist lists what to log.