Hi everyone! I've been lurking here for a bit, learning so much from all your posts on video workflows and AI tools. I finally took the plunge and created my first explainer video using HeyGen for a project at work. It's meant to show our sales team how to properly update lead sources in Salesforce to keep our reports clean.
I'm really happy with how it turned out overall! The avatar presenter looks natural, and the voiceover matched my script almost perfectly. I used a simple template and uploaded some screenshots of the Salesforce page to show the exact fields.
That said, I know I'm a total beginner at this. I’d love some feedback from people who have more experience. Specifically, I'm wondering:
* Is the pacing okay? I feel like the middle section might be a bit slow.
* The background music I chose is one of the free HeyGen tracks. Does it fit, or is it a little distracting?
* Are the screen annotations clear enough?
I’m sharing the link here. Any critiques—big or small—would be incredibly helpful as I try to make more of these for our internal training library. I’m especially keen on improving data quality through better training, so getting the video part right feels important!
Thanks!
I watched your video. You've got the basics down, but there are some fundamental issues that will undercut your goal of improving data quality.
First, you mentioned pacing in the middle. It's slow because you're showing static screenshots for too long. For a process walkthrough, you need screencast video of the actual clicks and typing. A screenshot with a circle animation isn't training, it's a slideshow. The annotation arrows are clear, but they point to a dead screen. That kills pacing and retention.
The background music is a minor issue. The real problem is the lack of a clear data story. You never show the *consequence*: what does a broken report look like? What's the cost of bad data? Show a corrupted Salesforce dashboard for two seconds *first*, then show the fix. That frames the "why" for your sales team, which is more important than the mechanics of clicking a field.
For internal training, skip the AI avatar. It adds an uncanny layer between your content and the viewer. A quick screencast with your real voice (even if it's not perfect) is more direct and trustworthy for procedural training. The time you spent on the avatar could have been spent showing a real-life example of a lead source being entered wrong and then corrected.
Your script matched, which is good, but the content structure is backwards. Start with the problem, then the solution.
—davidr
Totally agree about the lack of a "why." Showing the broken report first is key. It flips the whole video from a boring checklist to a real solution.
One caveat: real voiceover is great, but some teams are allergic to hearing their own manager's voice. An avatar can feel more neutral. The trick is nailing the script so it doesn't sound like a generic robot.
Demo or it didn't happen
The avatar looks "natural" because you're grading on a curve set by HeyGen. Compare it to any human presenter in a real training video. Uncanny valley, just polished enough to be creepy.
You asked about pacing and background music. That's like worrying about the paint color on a car with no engine. The problem is you built this on a platform designed for maximum friction to leave. Try exporting your script and assets. Bet you can't.
Your goal is improving data quality through training. A locked-down, single-vendor video tool is how you create a different kind of data quality problem. Vendor lock-in for a training library is a special kind of IT debt.
Your vendor is not your friend.
That lock-in point is a real gut check. I've seen teams build a whole internal knowledge base on a flashy platform, only to get a 300% price hike next year. You're stuck.
But the friction to leave isn't just about export formats. It's about maintenance. That video is tied to a specific Salesforce UI. When the layout changes in six months, you can't just re-record a 30-second clip. You have to rebuild the whole thing in HeyGen because you can't edit the original screen recording. That's the debt.
Sometimes the creepy avatar is the compromise to get a video made at all. The real failure is building a process that depends on it forever.
Run it yourself.