HeyGen Video Agent when somebody has to say it out loud
Every other video engine generates a scene and the sound around it. This one generates a person speaking — in Polish, with the words you give them, and with a mouth shaped to those words. That is a different kind of material and a different sort of engine: not a text-to-video model but an agent that composes the script into a shot itself.
Rendered by this model
Specification
- Built by
- HeyGen
- Starts from
- a written brief
- Aspect ratios
- 9:16 · 16:9
- Clip length
- 15 / 30 / 60 s
- Sound
- built in, always on
- Reference photos
- none
- Render time
- about 240s
- Cost
- from 5 to 20 credits per clip
What it is good for
- Polish, not a Polish accent
- The model defaults to English, so the language has to be stated in capitals in the brief and repeated. We do that for you and verify the result by transcription before the material comes back.
- Up to sixty seconds
- Six times longer than any other engine on the roster. Enough for a full statement: the problem, the product, the reason to buy.
- A mouth under the words
- Text-to-video models animate speech with nothing beneath it — the mouth moves but says nothing. Here the track and the picture are made together, so the material survives being watched with sound on.
When to pick something else
It renders the longest of anything here, about four minutes, and you cannot give it your own photo or product — the presenter comes from its library and the product has to enter in the edit. It takes only vertical and landscape. It is also not an engine for a product shot: for a bottle on stone take Veo, Seedance or Wan, because every frame here has a person in it.
See every modelHow to brief this model
The things this engine handles differently from the rest. The habits that hold for every model are collected separately, in the help centre.
- State the language in capitals and repeat it. This agent defaults to English regardless of the language the brief itself is written in.
- Dictate the lines word for word. Left with a general topic it invents a script, and in it brand names and sales figures nobody can stand behind.
- Match the text to the length. Fifteen seconds is about forty words of Polish — a longer script the presenter will rush to fit.
- Do not expect the product on screen. This engine takes no photograph of yours; the clip shows a person speaking, so order the product shot from Veo, Seedance or Wan and cut the two together.
- Plan for the wait. Rendering takes about four minutes, the longest of any video engine here, so it is not a model for reviewing versions.
Where this model works
Questions about this model
Other models
- Veo 3.1 FastThe default video engine. The cheapest second among the ones that make sound.
- Veo 3.1The best picture and camera move on the roster. And the priciest second.
- Kling 3 ProHands and objects in motion. For unboxing and for showing a product off.
- Seedance 2.5Nine reference photos. The only engine that holds your product through the shot.
- Seedance 2.0The Seedance picture at two thirds of the price. For clips up to eight seconds.
- Wan 3.0 PrimeTexture up close: cream, fabric, liquid. Five ratios, two credits a clip.
- Gemini Omni FlashThe cheapest clip whose sound cannot be switched off.
- MiniMax H3 MaxSlow motion and moving material. The highest resolution alongside LTX.
- Grok Imagine 1.5A strong grade and night scenes. For ads that are meant to be loud.
- Happy HorseFood, flowers, interiors. Natural light and a calm frame.
- LTX 2.3The only 1080p on the roster. And one of the two cheapest seconds.
- PixVerse V6One credit a clip. The cheapest video second in the whole studio.
Try this model on your own product
Sign up without a card and get 40 credits. Switching model is one click, so you can send the same brief through several of them and keep the best frame.