HeyGen Video 1 is now available on GPTunneL. The model creates short videos from a description, an image or references, with sound generated alongside the picture. If you need a product clip, an advertising scene or footage for a presentation, it is another option to evaluate for your workflow.
HeyGen introduced the model on September 30. Here is what the documentation says it can do, how I would approach a first generation, and what generation costs in GPTunneL. Prices were checked on October 1, 2026. GPTunneL generation starts at $0.02 per second with the current 50% discount.
What HeyGen Video 1 can do
Many people know HeyGen for its talking avatars. Video 1 generates an entire scene: objects, surroundings, camera movement and sound. The developer's catalog describes it as built on MiniMax H3 and further trained by HeyGen.
There are three ways to start:
- Text. Describe what happens in the shot, how the camera moves and what can be heard. Use this when you do not have an initial image.
- A first frame. Supply an image and describe the movement. Useful when you already have a product photo or illustration.
- References. Supply images, video or audio to guide generation. References let you show the intended scene and motion instead of describing everything in words.
A clip lasts 5–15 seconds, with 480p and 768p output. Landscape, square and portrait formats are supported. These parameters are documented by Runware and OpenRouter. A longer video requires several scenes and editing.
Sound is generated with the picture: you can describe speech, room ambience, footsteps or passing traffic. Listen to this arriving train example, where the audio is part of the action.
A HeyGen demo from its official catalog, not a generation from our own testing. Enable sound in the player.
What to try it for
I would begin with a short scene whose result is easy to assess. For a product listing, use one object, a quiet background and a slow camera push. This makes it easier to spot changes to the product's shape, color or important details.
For an ad creative, try variations of the same scene with different lighting, surroundings or angles. The first batch helps you choose a direction. Add headlines, prices and logos during editing so they do not depend on the model rendering letters correctly.
For an interior presentation, try a short camera move through a room. Check whether furniture changes shape or extra objects appear. The example below shows one of the developer's suggested uses.
HeyGen's “Real estate” text-to-video demo. Both clips come from the official collection.
For an accurate app walkthrough, I would still use a screen recording. A generated scene can work as an introduction, but button positions and interface text in a tutorial must match the actual product.
How much generation costs
HeyGen Video 1 currently has a 50% discount on GPTunneL. Platform catalog rates checked on October 1, 2026:
| Resolution | Discounted, per second | Standard, per second | 10 seconds, discounted |
|---|---|---|---|
| 480p | $0.02 | $0.04 | $0.20 |
| 768p | $0.03 | $0.06 | $0.30 |
For example, 10 seconds from text at 480p costs $0.20, or $0.30 at 768p. Five ten-second variations at 768p cost $1.50 with the current discount. Without the discount, those amounts are $0.40, $0.60 and $3.00 respectively.
For a first test, try 5 seconds at 480p for $0.10. If the idea works, try a longer scene or 768p. Budget for several attempts: not every result will make the final edit.
These prices were checked against GPTunneL's text-to-video calculation. Before generating, especially when using references, check the final cost on the model page: it accounts for your settings and the current discount. Current rates are also listed in the HeyGen pricing section.
Prepare your first prompt
Start with one object and one action. Specify lighting, camera movement and sound. Avoid packing several locations and a full storyline into a single short clip.
Here is a suggested prompt to adapt for your first test:
Close-up of a ceramic cup on a light wooden table beside a window. Steam slowly rises from hot coffee. The camera gently pushes toward the cup in a single continuous shot. Soft morning light, natural colors. Quiet room ambience and birds outside the window. No music, speech, lettering or logos.
When working from a photo, focus the prompt on what should move, such as the camera and steam. Do not describe a different material or shape if you want to preserve the original object.
After generation, check that the object stays consistent, motion has no jumps and sound matches the action. Change one condition at a time so you can tell which adjustment helped. You can start with a short 480p clip and repeat the chosen idea at 768p. A new generation may look different: this is not simply upscaling the existing file.
Try it in GPTunneL
Open HeyGen Video 1 in GPTunneL and select an available mode, duration and resolution. Check the final price before generating. An integration may not expose every API capability at the same time.
Choose a photo, write a prompt and decide what would make the result usable. To compare options, explore the available video models. Our AI video creation guide can help you prepare your first scene.