How to Automate Personalized Video Outreach with HubSpot, ChatGPT, and HeyGen
Cold outreach has a reply problem. Plain-text emails now land somewhere between 1% and 5% response rates, and most of them never get opened at all. Video changes that math. Personalized videos see roughly 16 times higher click-to-open rates and about 4.5 times more clicks than generic content, according to Close's research on video personalization. The catch has always been time. Recording a custom video for every single lead doesn't scale past a handful of prospects a day.
That's the problem this workflow solves. Instead of a rep manually researching a lead, writing a message, and recording a video, you can chain four tools together so the entire sequence fires on its own the moment a new lead lands in your CRM.
Why Combine HubSpot, ChatGPT, HeyGen, and Gmail
Each tool in this stack does one job well, and none of them talk to each other out of the box. That's where an automation layer comes in.
HubSpot holds the lead data and triggers the workflow the moment a new contact is created or a form is submitted.
ChatGPT turns raw lead data (job title, company, industry, recent activity) into a short, personalized script.
HeyGen takes that script and generates an AI avatar video that says it out loud, using a face and voice the recipient hasn't seen a thousand times before.
Gmail sends the finished video, embedded or linked, in a follow-up email that looks like it came from a human who did their homework.
On their own, these are four separate logins and four separate manual steps. Connected through an automation platform, they become one continuous pipeline that runs whether or not anyone on the sales team is at their desk.
Sales teams that already automate parts of their outreach report meaningful time savings. HubSpot's own research found that sales reps using AI and automation tools save over two hours a day on tasks like outreach creation, data entry, and follow-ups, according to HubSpot's Smarter Selling with AI report. Video outreach is one of the highest-leverage places to point that saved time, because it's also one of the hardest things to do manually at volume.
The Workflow, Step by Step
Here's how the pieces fit together in practice.
1. A new lead enters HubSpot
This can be a form fill, a demo request, or a contact manually added by a rep. The trigger fires as soon as the lead record is created or reaches a specific lifecycle stage, so nothing sits in a queue waiting for someone to notice it.
2. ChatGPT researches and drafts the message
The lead's name, company, title, and any notes get passed to ChatGPT with a prompt that asks it to write a short, natural-sounding script referencing something specific to that person or company. This is where generic outreach usually falls apart. A script that says "I saw you're the VP of Marketing at [Company]" instead of just "Hi there" is the difference between a video that gets watched and one that gets deleted.
3. HeyGen turns the script into a video
The script gets sent to HeyGen, which generates a talking avatar video reading it aloud. You can set a consistent presenter (a rep's likeness or a branded avatar) so every video looks and sounds like it's coming from the same person, just with a different message each time.
4. Gmail sends the video automatically
Once HeyGen finishes rendering, the video link or thumbnail gets dropped into an email template and sent through Gmail, often with a follow-up task logged back in HubSpot so the rep knows the touch happened and can pick up the conversation if the lead replies.
The whole sequence, from lead capture to inbox, can run in minutes without a single manual step.
What You Need to Set This Up
You don't need custom code to wire these four tools together. What you do need is a way to pass data between them, since HubSpot, ChatGPT, HeyGen, and Gmail were never built to share information directly. This is normally handled with a workflow automation tool that connects to all four through their APIs, listens for the HubSpot trigger, and passes the output of each step into the next.
If you're building this yourself, look for a platform with pre-built connections to all four apps rather than one that only handles a couple of them. viaSocket's integration coverage spans thousands of apps including HubSpot, OpenAI's models, HeyGen, and Gmail, so you're not stitching together separate tools for each connection point.
Comparing Manual vs. Automated Outreach
Step | Manual Process | Automated Workflow |
|---|---|---|
Lead research | Rep checks LinkedIn, company site, notes | ChatGPT pulls context from HubSpot fields instantly |
Message writing | Rep drafts by hand, 10-15 minutes per lead | ChatGPT generates a draft in seconds |
Video creation | Rep records on camera, re-records mistakes | HeyGen renders automatically from the script |
Sending | Rep manually attaches and sends | Gmail sends as soon as the video is ready |
Time per lead | 20-30 minutes | Under 5 minutes, no rep time required |
Scale | Limited to a few leads per day | Hundreds of leads without added headcount |
The gap widens fast once lead volume grows. A rep who can personally handle 10 quality video touches a day hits a wall; the automated version doesn't.
Getting the Personalization Right
Automation only helps if the output still feels personal. A few things make the difference between a video that reads as thoughtful and one that reads as obviously templated:
Feed ChatGPT specific fields, not just a name. Job title, industry, and any deal notes give it something real to reference.
Keep the script under 45 seconds. Short videos get watched all the way through far more often than long ones.
Use a subject line that names the video and the person, something like "A 40-second video for Sarah," since naming the format tends to lift open rates.
Review a sample batch weekly. ChatGPT can drift toward generic phrasing if the prompt isn't tight, so spot-check the scripts before they go out.
If you want to see the full workflow mapped out visually before building it, viaSocket's library of automation templates includes similar multi-app sequences you can adapt rather than starting from a blank canvas.
Common Pitfalls to Avoid
Teams that try this and get lukewarm results usually run into one of three problems. First, they skip the research step and let ChatGPT write from the name alone, which produces the same flat message HeyGen just happens to read aloud instead of type. Second, they use a single generic HeyGen avatar for every rep, which breaks the illusion that a specific person made this specific video. Third, they don't log the send back into HubSpot, so reps have no idea which leads already got a video and end up sending a second one by accident.
None of these are automation failures. They're setup mistakes, fixable by tightening the prompt, assigning avatars per rep, and logging the send back to the contact record.
FAQ
Do I need to code this workflow myself?
No. HubSpot, ChatGPT, HeyGen, and Gmail all expose APIs, but a no-code automation platform can connect them without custom scripts. You're mainly configuring triggers, prompts, and templates rather than writing integration code.
How long does a lead wait between form fill and receiving the video?
That depends on HeyGen's render time for the specific video length and avatar, but most short clips (under a minute) finish rendering within a few minutes, so the full sequence from lead capture to sent email usually completes in well under ten.
Will this work for outbound prospecting, not just inbound leads?
Yes. The trigger doesn't have to be a form fill. It can be a rep adding a contact manually, a list import, or a lead reaching a certain score, as long as HubSpot has the fields ChatGPT needs to personalize the script.
Is this only useful for sales teams?
Sales is the most common use case, but the same pattern works for customer onboarding, renewal check-ins, or recruiting outreach. Anywhere a personal-feeling video beats a form email, the workflow applies.
The Payoff
None of these four tools were designed to work together, but connected through automation, they turn a task that used to eat a rep's entire morning into something that happens in the background. The lead shows up, the research happens, the video renders, the email sends, and the rep's first real involvement is reading a reply. That's the actual promise of AI sales automation: not replacing the conversation, just getting a real, personalized one in front of more people than a person alone ever could.
