HeyGen: Avatar IV

HeyGen: Avatar IV is a video generation model from HeyGen, a company known for its work on AI avatars and talking-head video. It sits in the video category and is built specifically for animating people, rather than for general-purpose footage generation. The model was released by the provider on 24 August 2026.

Avatar IV is an image-to-video model. It takes a single photo and animates it into an expressive, lip-synced talking-head video. Its inputs cover text, image and audio, and its only output is video. Rather than only matching mouth shapes to words, it interprets the vocal performance, so the resulting motion reflects how something is said and not just which words are spoken. Output can be produced at 720p or 1080p, in 16:9, 9:16 or 1:1 aspect ratios. Audio generation is not part of the model's output, so the audio side of a clip is not produced by Avatar IV itself. No context length is specified for this model.

The practical uses follow from that combination. A still portrait plus a script or a voice track is enough to produce a presenter-style clip, which suits training and explainer content, product walkthroughs, and localized versions of the same speaker for different audiences. The three aspect ratios make it straightforward to target different placements: 16:9 for standard video players and web embeds, 9:16 for vertical feeds, and 1:1 for square formats. Because the input can be a single photo, teams can build a consistent on-screen presenter without filming new footage each time, and because the model responds to vocal delivery, the same image can carry very different tones depending on the audio it is given.

You call Avatar IV through api14's OpenAI-compatible API using the model id "heygen/avatar-iv". The request shape follows the same conventions as other models on the gateway, so existing OpenAI client libraries and tooling work without changes. The model accepts text, image and audio inputs and returns video. No additional request parameters are listed as supported for this model, so keep requests to the standard fields and avoid assuming provider-specific options such as duration or frame-image controls will be accepted.

There are a few things to keep in mind when choosing it. Avatar IV is a talking-head model, so it is not the right fit for scenes without a person, for general b-roll, or for work that needs generated audio, since audio generation is disabled. If your task is text generation, reasoning or image understanding rather than producing video from a portrait, a text or vision model on api14 will serve you better.