HeyGen Video 1 Explained: API, Native Audio and 2K Limits

HeyGen has launched Video 1, its first general-purpose video model, in its product and API. It generates short scenes with dialogue and sound effects from text, images or references. For teams already making avatar videos, the useful addition is a way to create the surrounding footage in the same workspace.

The launch appeared in HeyGen's October 21 blog listing and official announcement post. HeyGen advertises fast generation, but the more practical questions are which endpoint to use and what changes when you request higher resolution.

What HeyGen Video generates

The product page describes clips between five and fifteen seconds, with picture and audio generated together. Instead of choosing a presenter and supplying a script, you describe a scene, its subject and its sound.

Text-to-video starts from that description. Image-to-video makes the supplied image the opening frame. Reference-to-video uses supplied material to guide a new scene, rather than requiring the result to start on one particular still.

HeyGen supports up to nine image references, three videos and three audio clips, with twelve references in total. That total matters: the individual maxima are not additive.

This is an addition to the avatar workflow, not a replacement for it. A presenter can deliver a product explanation while generated clips provide the establishing shot or visual cutaway. The distinction helps avoid asking a scene generator to do the job of a tightly controlled presenter system.

API access and resolution limits

The developer documentation identifies the model as heygen-video-1, built on MiniMax H3 and post-trained by HeyGen.

Create a job with POST /v3/models/videos, then retrieve its status and output with GET /v3/models/videos/{video_id}. This is separate from the avatar-rendering route, POST /v3/videos. The model route requires a paid API key.

Setting	Documented support
Duration	5 to 15 seconds, or automatic selection
Resolution	480p, 768p, 1080p and 2K
Higher-resolution billing	1080p and 2K cost three times the 768p rate
Text/reference aspect ratios at 1080p and 2K	16:9 or 9:16

The announcement's short pricing claim does not spell out a complete resolution-by-resolution tariff. Check the account's API pricing before calculating a production budget. In particular, do not assume the advertised entry rate applies to 2K output.

That multiplier changes the sensible evaluation process. Test composition and whether the action works at the lower resolution first. A sharper version of a failed shot still needs to be regenerated.

Where generated scenes need review

HeyGen's own guidance favors contained shots with one subject and limited camera movement. It flags close hand work and long text as weak areas, so a clip demonstrating a physical procedure deserves more scrutiny than a simple atmospheric cutaway.

For a product launch, the reference image is useful input, but it is not a guarantee that every visible detail survives correctly. Review the actual result for labels, shape and the action being shown. Add important specifications as ordinary edited text rather than trusting generated lettering to carry the information.

Native audio also changes the review workload. A visually acceptable scene may still have an unwanted sound, awkward speech or a mismatch with the rest of the edit. Generating the two together removes a separate creation step, not the need to listen.

The worthwhile test is therefore cost per usable shot, including failed attempts and review time. Video 1 makes scene generation available inside an existing business-video workflow. Whether that simplifies production depends on how often its output can move directly into the edit.

I'm Paras. I work at the intersection of AI, engineering and distribution.

I help technical teams turn AI launches and product capabilities into clear stories that people understand and share.

Book a 30-minute call.