HeyGen

Your assistant can list the videos in the account, read the processing status and metadata for any one of them, and pull the remaining quota alongside your account details. That covers both halves of the question people usually ask separately: what exists, and whether there is enough left in the plan to make the next batch. It reads only.

What you end up with. An inventory of your videos with their status, and the remaining quota stated next to it.

Start with it bringing you the information and you deciding.

Never without. Treat quota as the figure the account reports right now, not as a forecast. It moves as other people in the account render, and a batch job can consume it while you are reading the answer.

Built on: List videos / Retrieve video status details / Retrieve user remaining quota / Get current user information

This is what HeyGen templates are for, and the connector covers the loop. It can list the templates in the account, read the variables a specific template exposes, then generate one video per set of values. Two practical notes. The template action supports a test mode, which is the right way to see the output before spending credits on the full run. And rendering is asynchronous, usually 30 to 180 seconds or longer per video, so the assistant submits and then polls rather than waiting.

What you end up with. One rendered video per row of your list, each with its video ID and status, plus a named list of any that failed.

Start with it doing the work and checking with you.

Never without. Run one in test mode and watch it before releasing the batch. A variable that renders slightly wrong is a typo in one video and the same typo in forty, and every one of them has your face on it.

Built on: Retrieve all templates / Get template by id / Retrieve template details v3 / Generate video template with variables / Retrieve video status details

It can list the avatars available to the account and read the details of a specific one, list the voices, and then generate a video from your script with the avatar and voice you chose. The output is a video ID that it polls until rendering completes. The part worth being deliberate about is the choice of avatar and voice, because those are what the audience actually reacts to and they are the cheapest thing to get wrong.

What you end up with. A rendered video of your script, with the avatar and voice named so the next one matches.

Start with it doing the work and checking with you.

Never without. Never generate a video using a likeness or a voice modeled on a real person without that person's written agreement to this specific use. A synthetic person saying something they did not say is the risk this whole category carries, and it does not become smaller because the render was easy.

Built on: List avatars endpoint / Retrieve avatar details / List available voices / Generate video with inputs / Retrieve video status details

HeyGen can translate a video into another language, either the whole video or the audio track alone. The assistant can first read which target languages are supported, including dialects and locales, submit the translation, and then poll the job until it reports complete or errored. What comes back is a separate video, so the original is untouched.

What you end up with. A translated version of the video as its own asset, with the source language version left as it was.

Start with it doing the work and checking with you.

Never without. Never publish a translation nobody who speaks the language has watched. Machine translation gets product names, legal wording and units of currency wrong in ways that read as careless to the audience you made it for.

Built on: Retrieve video translation target languages / Post video translate request / Retrieve video translation by id

Rather than rendering a full video to find out the voice is wrong, it can list the available voices and locales and generate a short audio sample of your actual script text. There is a straightforward text-to-speech action for this, and a separate short preview action that HeyGen lists as Enterprise only, in beta, and consuming API credits. Use the first unless you know you have the second.

What you end up with. Audio samples of your own script in the shortlisted voices, so the voice decision is made before the video budget is spent.

Start with it doing the work and checking with you.

Never without. Sample the awkward sentence, not the easy one. Voices differ most on names, acronyms and numbers, and those are exactly the lines that will be in the finished video.

Built on: List available voices / List all locales for voices / List brand voices / Generate text-to-speech audio / Generate voice audio preview

It can generate a public shareable URL for a specific video. HeyGen is explicit about what that means: anyone with the link can watch it, no HeyGen account and no authentication required. Links may be time-limited, and asking again produces a fresh one when an old one expires. That makes it genuinely useful for a client review and genuinely unsuitable for anything confidential.

What you end up with. A working public link to the video that the recipient can open without an account.

Start with it doing the work and checking with you.

Never without. Never create a public link for a video containing client names, pricing, or anything unreleased. The URL bypasses authentication entirely, so it protects nothing once it has been forwarded, and you cannot tell who has it.

Built on: Retrieve video status details / Retrieve sharable video URL

It can list the folders in the account, create new ones, rename them, and move a folder to trash, which HeyGen describes as a soft delete with a matching restore action. It can also list and upload assets, the images, audio and video files that video generation draws on. Worth knowing before you start: videos are placed in a folder at the moment they are generated, by passing the folder ID, so organizing well is mostly a decision you make before rendering rather than after.

What you end up with. A folder structure that matches how you actually look for things, with the assets listed and the empty folders cleared.

Start with it doing the work and checking with you.

Never without. Check what is inside a folder before it goes to trash, and note the folder ID in case you need the restore action. Moving a folder takes everything in it out of where colleagues expect to find it, and the recovery only works if you can still say which folder it was.

Built on: List folders / Create a new folder / Update folder name / Move folder to trash / Restore folder from trash / List assets / Upload asset

It can delete a video by ID, covering both generated and translated videos, and delete an asset such as an uploaded image, audio file or video. Unlike the folder action, neither of these is described as recoverable. Run it in two moves: have it list what matches your rule, read that list yourself, then approve deletion of exactly those IDs.

What you end up with. The old drafts and uploaded source files removed, and a saved list of every video and asset deleted with its ID and title.

Start with it doing the work and checking with you.

Never without. Never approve a deletion described only as a rule, and download anything you might need again before it goes. HeyGen offers a restore path for a trashed folder and none for a deleted video or asset, and deleting the source asset means the video cannot be regenerated either.

Built on: List videos / List assets / Delete an avatar video / Delete asset