HeyGen Video generates picture and sound at $0.02 a second

HeyGen has shipped HeyGen Video, its first general-purpose video model, which turns a single prompt into a 5 to 15 second scene with its own sound. The company announced it in a blog post and a help article. The help article is dated September 30, 2026, and OpenRouter gives the same release date, while Tech Social Network dated the announcement to October 21, 2026.

HeyGen's argument is that most AI video is assembled from parts: one model draws the picture, another voices the lines, lip-sync stitches them together. HeyGen Video makes picture and sound in the same pass, so every clip arrives with its own dialogue, ambience and sound effects, and there is no separate lip-sync step.

The model is a departure for a company known for avatars. The model page says HeyGen Video is built on MiniMax H3 and post-trained by HeyGen. HeyGen suggests pairing it with Avatar V: open on a generated scene, hand the message to an avatar, cut back to a reference-to-video shot of the product.

Clips run 5 to 15 seconds and take three kinds of input: text, a first-frame image, or up to 12 reference files (9 images, 3 videos and 3 audio clips) that the prompt addresses by name. Output is an MP4 at 24 frames a second with stereo audio, in aspect ratios from 21:9 to 9:16 and resolutions of 480p, 768p, 1080p or 2K. Prompts can run to 32,000 characters, and setting a seed returns the same clip for the same prompt.

HeyGen bills by the second of output. Text to video and image to video cost $0.02 a second at 480p and $0.03 at 768p; reference to video costs $0.04 and $0.06, counting the seconds of any reference video as well. Full HD and 2K bill at three times the 768p rate, and a 50% launch discount runs through the end of October, putting 480p at $0.01 a second.

On price, HeyGen's own developer catalog lists its model at $0.03 a second at 768p with audio, against $0.168 for Kling 3.0 Pro and $0.40 for Veo 3.1 on the same page. Those are the maker's listed figures, not the rivals' own rates.

HeyGen also says the model ranked first in blind tests against Seedance 2.5, Seedance 2.0, H3 Max and Kling 3.0, from 4,800 blind votes on the Artificial Analysis Arena query set. The company does not name who ran the comparison and cites no outside test, so it stays a claim.

For creators, the appeal is a cheaper route to short shots with sound already inside them; the 15 second ceiling means long scenes take more than one generation.

HeyGen Video runs on paid API keys. Get one at app.heygen.com/developers/api, or reach the model through Creative Studio, OpenRouter, Runware and ComfyUI.