Nine cents for a 4-second AI video – that’s the headline that makes anyone who budgets for motion graphics sit up. ByteDance’s Seedance 1.5 delivers exactly that price point, but the trade-off is a model that still wrestles with basic physics.
My test prompt was deliberately simple: two guys playing basketball with a watermelon. The output was a 480p clip, four seconds long, with synchronized audio. The total charge was $0.09; strip the audio and the cost drops to roughly $0.045. Compared with other text-to-video services that charge dollars per second, the economics are striking.
Each 720p 5-second video with audio costs roughly $0.26. For other resolutions, 1 million video tokens with audio cost $2.4, and without audio the price is $1.2 per million tokens. Tokens are calculated as (height × width × FPS × duration) / 1024. Plugging in 1280 × 720 at 30 fps for five seconds yields about 135 000 tokens, which aligns with the $0.26 figure when multiplied by the $2.4 / M-token rate. This explains why the 480p, 4-second clip landed at nine cents.
Visually the clip holds up. Lighting is even, textures are clean, and the watermelon’s surface reflects light in a believable way. Audio sync is on point – the spoken lines line up with mouth movements, something many video generators still miss.
The physics, however, betray the model’s limitations. After the player launches the watermelon toward the hoop, the fruit disappears into the net, yet the player still has it. The scene violates object persistence and basic cause-and-effect. For a scenario that relies on a single ball-throw, the error is glaring; for a static talking-head or a product showcase, it would be invisible.
Language handling gave another clue about the model’s defaults. The first run produced dialogue in Mandarin, which I could not understand. Adding the phrase “they are Americans” forced English output. This suggests Seedance assumes a Chinese-centric model unless the prompt explicitly overrides it. International teams will need to be precise about language requirements.
Be aware that the model can get confused with multi-speaker dialogue. You should specify the full description of each person every time they speak. Instead of simply switching between names like ‘Bob’ and ‘Alice,’ your script should say, ‘Bob, the guy on the right, says this,’ and then, ‘Alice, the girl on the left, says this.’ You must repeat those identifiers throughout the script.
Here’s an example of a dialogue test that I ran that did much better with this approach (though it still isn’t perfect, as you can see).
When I ran it without this trick, it constantly swapped dialogue between the characters without caring about the prompt. With the trick, at least both characters said their own lines.
Where Seedance 1.5 Makes Sense
Because the price is so low, the model fits into a niche of rapid prototyping and inexpensive stock-footage generation. Marketing teams that need background motion for a splash screen, educators creating quick visual aids, or indie creators looking for a B-roll filler can all benefit. The key is to avoid scenes that demand realistic complex interaction between objects.
Practical Tips for Getting the Most Out of Seedance 1.5
- Keep the action simple. Stick to static cameras, single characters, and minimal object interaction.
- Specify language up front. Include a clear cue like “English dialogue” to avoid default Mandarin output.
- Budget per frame. At $0.09 for four seconds you’re effectively paying $0.0225 per second. Plan your clips accordingly.
- Use the output as a scaffold. You might need to export the video, then overlay graphics or replace problematic frames in a traditional editor.
For anyone who has been watching the AI video race, Seedance 1.5 is a reminder that the market is diversifying. Not every player is racing toward perfect realism; some are racing toward affordability and speed. The model’s physics issue is a genuine limitation, but it’s also a clear signal of where research effort is still needed.
If you need a quick visual prototype and can tolerate a little physics-magic, nine cents is a price you can test without worrying about budget overruns. For anything that requires believable interaction, you’ll have to use premium models or combine Seedance’s output with traditional VFX pipelines.
Bottom line: Seedance 1.5 delivers cheap, decent-looking video for simple shots.
