Connect HeyGen to Your CRM with Make (Formerly Integromat)

What You Need Before You Start

You need a Make account on the Core plan or above (the Free plan’s 1,000 operations per month is not enough for live video rendering), a HeyGen account on Creator or above with API access enabled, and a CRM that Make can connect to. This guide uses HubSpot for the trigger and write-back modules, with notes on swapping to other CRMs.

Before building the scenario, go into your HeyGen dashboard and confirm two things: the avatar ID you plan to use is active, and the template variables match what you will pull from the CRM. Mismatched variable names produce blank fields in the output video - no error, just silent bad output.

Step 1: Set Up the CRM Trigger Module

Create a new scenario in Make. For HubSpot, add the HubSpot - Watch Deals module as the trigger. Set it to watch for stage changes and select the pipeline and stage you want to act on. Connect your HubSpot account when prompted.

If your CRM is not in the Make app library, use a Webhooks - Custom Webhook module instead. Copy the generated webhook URL, add it to your CRM’s outbound webhook settings, and Make will receive the payload and map the fields. This approach works for Pipedrive, ActiveCampaign, Close, and any CRM that supports outbound webhooks.

After the trigger, add a Router module with two paths: one for records that have all required personalization fields populated, one for records that are missing fields. Route incomplete records to a Slack notification or an email alert rather than passing them into the render path.

Step 2: Call HeyGen via the HTTP Module

Add an HTTP - Make a Request module. Set Method to POST and URL to https://api.heygen.com/v2/video/generate. Under Headers add X-Api-Key with your HeyGen API key. Set Body Type to Raw and Content Type to JSON. The body structure:

The {{1.firstname}} notation is Make’s field mapping syntax - replace the field names with whatever your trigger module exposes. Enable Parse Response in the module settings so Make automatically maps the response JSON. The value you need from the response is data.video_id. Keep test: true until you have verified output on two or three real records.

Step 3: Handle Async Rendering in Make

HeyGen renders do not complete instantly. After the render POST, add a Sleep module set to 60 seconds, then an HTTP module that polls the status endpoint. Use a Repeater module to loop up to 10 times:

1. Add a Repeater module. Set maximum repetitions to 10.
2. Inside the repeater, add a Sleep module - 30 seconds.
3. Add an HTTP - Make a Request module: GET https://api.heygen.com/v1/video_status.get?video_id={"{{"}video_id{"}}"} with the same X-Api-Key header.
4. Add a Break module after the HTTP module with the condition: data.status equals completed or failed.

After the repeater exits, add a Router that checks the final status value. Route completed to the write-back path and failed or still-processing to the error path.

One limitation to know upfront: Make’s Sleep module holds the scenario execution open. At higher volumes this fills your active execution slots. If you are running more than 20 renders concurrently, switch to HeyGen’s webhook-based completion - register via POST /v1/webhook/endpoint.add and point the callback URL at a second Make scenario that handles the write-back.

Step 5: Write the Video URL Back to Your CRM

After the successful completion branch of the router, add a HubSpot - Update a Contact Record module (or the deal equivalent). Map the data.video_url field from the status response to a custom CRM property you created for the video link.

If your CRM does not have a native Make module for the write-back, use an HTTP - Make a Request module with a PATCH or POST to the CRM’s REST API. For Pipedrive this is POST /v1/notes with the deal_id and the video URL in the content field. For Salesforce it is a PATCH to /services/data/vXX.0/sobjects/Contact/ID.

Include the render date in whatever field or note you write. HeyGen URLs expire after 7 days. If a rep opens the CRM record two weeks later and clicks a dead link, they will assume the integration is broken - which is worse than no integration.

The Ceiling: Where Make Stops Working Well

Three things constrain Make for HeyGen at volume, and they compound each other.

Per-operation pricing at volume. A single render run costs 5 to 8 Make operations. At 50 videos a day that is roughly 10,000 to 12,000 operations per month - the entire Core plan’s allowance, spent on one scenario. The Pro plan at 10,000 operations is similarly tight. At this point you are paying Make more per render than the HeyGen API costs.

Scenario timeouts on slow renders. Make scenarios have a 40-minute hard timeout. A single render will not hit it, but if HeyGen is slow or your polling loop runs the full 10 iterations, you are sitting at 5 to 8 minutes per execution. Concurrent renders on a busy day stack those timeouts and start killing executions.

No managed retry queue. Make’s error handling is per-scenario and stateless. If a render fails and your error handler is not perfectly configured, the job is gone. Custom code with a database-backed queue gives you dead-letter storage, retry visibility, and aggregate failure-rate monitoring - none of which Make can provide natively.

When to Move to Custom Code

Move off Make when any of these are true: you are generating more than 50 videos per day, you need a delivery SLA with guaranteed retry and dead-letter queuing, you need aggregate monitoring of render success rates, or your Make operation bill is approaching the cost of a small server or serverless function.

The migration is not a rewrite. The HeyGen API calls are identical. What changes is the orchestration layer - a queue (SQS, BullMQ, or similar), a worker process that handles the polling loop, and a monitoring dashboard. Make is a good place to validate the integration before investing in that infrastructure.