Microsoft Word doc/docx to PDF conversion on AWS Lambda using Node.js
Branch: master
Clone or download
Type Name Latest commit message Commit time
Failed to load latest commit information.
lambda v1.1.0: support invocation from AWS Step Function Feb 18, 2019
LICENSE initial release Feb 1, 2019 Merge branch 'master' of… Feb 18, 2019



Convert a Word document (.doc or .docx) in a source S3 bucket to PDF, saving the PDF to a destination S3 bucket.

This Lambda can be invoked from an AWS Step Function, or in response to an S3 "created" event. It could easily be modified to support other triggers. You probably still want to use S3 buckets, to workaround any limits on request/response size.

Thanks to Lambda's concurrency, this approach is well-suited to variable bulk/batch higher-volume conversion workloads.

This app uses Native Documents' docx-wasm to perform the conversion. It does not use LibreOffice etc.

Installation and Getting Started

Direct link to deploy or search for "docx-to-pdf" in the Serverless Application Repository.

Please double check you are in the AWS region you intend; this needs to be the same region as the bucket which will contain the Word documents you wish to convert.

After you click "Deploy" (bottom right corner), you'll need to wait a minute or so as CloudFormation creates resources. When this is complete, you should see a green tick saying "Your application has been deployed"

Now go into the function: Lambda > Functions then configure an S3 trigger (or your step function), environment variables and execution role as explained below.


This function can respond to S3 ObjectCreated events. In this case, the output PDF is the input key + .pdf.

To configure the trigger, in "Designer > Add triggers", click "S3". The "Configure triggers" dialog appears.

  • Select a bucket (any time a docx is added to this bucket, the function will run)

  • Verify that "all object create events" is selected (or choose PUT POST or COPY)

Click "Add" (bottom right), then "Save" (top right).

AWS Step Function

If you want to use docx-to-pdf in an AWS Step Function, you don't need the above trigger.

Instead, add a state of type task. Here is a working demo step function:

  "Comment": "docx-to-pdf conversion step",
  "StartAt": "DocxToPdf",
  "States": {
    "DocxToPdf": {
      "Type": "Task",
      "Resource": "arn:aws:lambda:us-west-2:992364999735:function:cloud9-serverlessrepo-docx-to-pdf-S3Fn-Z2UN6XIPAZ8B",
  "Parameters": {
       "source_bucket.$": "$.source_bucket",
       "source_key.$": "$.source_key",
       "target_bucket.$": "$.target_bucket",
       "target_key.$": "$.target_key"    
      "End": true

Replace the Resource value with the ARN for your Lambda (shown in the Lambda Designer, top right)


This application uses Native Documents docx-wasm library to perform the conversion.

So you need a ND_DEV_ID, ND_DEV_SECRET pair (or ND_LICENSE_URL) to use it. We have a generous free tier, you can get your keys at

Now set these as environment vars in the Lambda console, as described below.

Environment Variables

On the same screen in the Lambda Management Console for this function, scroll down to "Environment Variables":

If you are using an S3 trigger (as opposed to a step function), you also need:

  • S3_BUCKET_OUTPUT: the name of the S3 bucket to which the PDF will be saved (if blank, it should write to the input event bucket)

Execution role

Choose or create an execution role. In IAM, confirm that role's policies includes CloudWatch Logs permissions, and also:

            "Effect": "Allow",
            "Action": [
            "Resource": "arn:aws:s3:::*"

Without GetObject permission on the triggering bucket and PutObject permission on the output bucket, you'll get Access Denied errors.

Confirm installation is successful


If you configured the S3 trigger, you can try it, by copying a Word document (doc or docx) into the S3 bucket you have set the trigger on.

To verify it works, look for a PDF in your output bucket, or check the logs in cloudwatch

AWS Step Function

If you used the sample step function, you can "start execution", then for the input, use something like:

   "source_bucket": "MyDocxBucket",
   "source_key": "path/to/my/docx",
   "target_bucket": "MyPDFBucket",
   "target_key": "path/to/my/PDF"

substituting your own values.

To verify it works, check the execution status and/or event history, or look for a PDF in your target bucket at target key. You can also check the logs in cloudwatch


By default, this application logs to CloudWatch at "debug" level. You can turn debug level logging off by setting environment variable DEPLOY_ENV=PROD.

Sizing notes

Conversion is processor bound. Lambda allocates processor in proportion to memory.

Experience suggests 2048MB results in enough processor that conversion is both faster and cheaper (on Lambda January 2019 pricing) than lesser amounts. Note: Conversion uses a minimum of around 500MB RAM, and WebAssembly under Node can use a maximum of 2GB RAM.

Source code

To use the source code, clone the GitHub repo.

In the lambda dir, run npm install

Now you can use your favourite lambda development environment...


If you are having trouble, please check the CloudWatch logs.

For each conversion job, "RESULT: Success" or "RESULT: Failed" will be logged.

Checking for conversion failures:

  • if you are using an S3 trigger, then in the case of "RESULT: Failed", the source document will be copied to the desitnation bucket under the key "BROKEN". This makes it easy to check for conversion failures. If there is a failure, the reason for that failure will appear in the CloudWatch logs.

  • if you are using the demo step function, the in the Step Functions Management Console, you can look for executions with status "failed". If you click into one of these, you'll see the reason in the execution event history, under "ExecutionFailed".

Here is what may be going wrong:

  • Lambda "Task timed out" or out of memory ("Process exited before completing request"). You can increase Memory and Timeout parameters in the Lambda Management console.

  • Network error A network connection is required to validate your ND_DEV_ID, ND_DEV_SECRET (but not to perform the actual conversion)

  • TokenValidationError mean an invalid ND_DEV_ID, ND_DEV_SECRET pair. Did you get these from and declare them in the Lambda console?

  • OperationFailedError Mainly thrown when loading a document. Is this a Word (doc/docx) document? Please verify it opens correctly in Microsoft Word, or drag/drop it to If you continue to have issues, please try a simple "Hello World" document.

  • EngineExceptionError An exception in the document engine occured. Please let us know about this!

Getting Help

If you continue to have problems, please ask a question on StackOverflow, using tags #docx-wasm, #ms-word, #pdf, #aws-lambda, and #amazon-s3 or #aws-step-functions as appropriate, or post an issue on GitHub.