Stable Video Diffusion Environmental Impact
Open-source video generation model
One query, in context
Energy, carbon and water for a single use. One ~4-second video clip (14-25 frames)
About 1.4× a kettle boil.
About 29× 10 m by car and 29% of a km by car.
About 1.1× a glass of water.
- Where it runsSelf-hosted (varies)
- Grid carbon intensity458 g CO₂/kWh · 32% renewable
- Compared with a Google search500x more than a search (0.3 Wh)
- Water~4 queries to fill 1 litre
Where Stable Video Diffusion sits
#181 of 199 by energy per query. Heavier than 90% of all models. Lit dots are the same kind of model; click any dot to open it.
Want something lighter?
Stable Video Diffusion is already among the lighter well-known options for this kind of work, so there isn’t a clearly lighter alternative to suggest.
What a month of use adds up to
Pick how much you use it. Based on one quick use at a time, five days a week.
- Energy
- 3.25 kWh
- Carbon
- 1.30 kg
- Water
- 6.06 L
A month of 1 clips a day, 5 days a week. The energy is about 30× a kettle boil and 44% of a UK home's day.
Work out your own →- Energy
- 6.50 kWh
- Carbon
- 2.60 kg
- Water
- 12.12 L
A month of 2 clips a day, 5 days a week. The energy is about 59× a kettle boil and 88% of a UK home's day.
Work out your own →- Energy
- 25.98 kWh
- Carbon
- 10.39 kg
- Water
- 48.50 L
A month of 8 clips a day, 5 days a week. The energy is about 3.5× a UK home's day and 12% of a UK home's month.
Work out your own →How does Stable Video Diffusion compare?
Next to similar models. Bars use a log scale, so each step is a big jump.
Detailed breakdown
Energy
Stable Video Diffusion generates short video clips (14-25 frames) from image inputs. At ~150 Wh per clip, it is among the more efficient video models due to its relatively small parameter count (1.5B). Being open-source, it can be self-hosted and optimised for specific hardware.
Power source & carbon
Open-source model that can be deployed on any GPU infrastructure. A single H100 GPU can generate clips, making it accessible for self-hosting.
Water use
At ~280 mL per clip. Self-hosted on consumer hardware, water consumption is zero. Cloud-hosted, it depends on the provider's cooling infrastructure.
About Stable Video Diffusion
Stable Video Diffusion sits at the high end of AI energy consumption at 150.0 Wh per generating one short video clip — over 500x what a Google search uses. Stability AI's video generation model, released in November 21, 2023, open-source video generation model. The extreme energy cost reflects the computational complexity of video generation: generating coherent frames across both space and time.
These figures are estimates derived from hardware specifications and API benchmarks — Stability AI has not published official energy data for Stable Video Diffusion. Actual consumption may vary significantly depending on batching, quantisation, and infrastructure optimisations that we cannot observe from outside.
Stable Video Diffusion in context
Key insights
- Open-source weights — can be self-hosted on infrastructure you control
The Stability AI Stable family
One chart per tier, on a shared time axis. Hover a dot for details; click to open that version.
What does your Stable Video Diffusion usage cost the planet?
Use the calculator to estimate your personal footprint based on how often you use Stable Video Diffusion.
Frequently asked questions
How much energy does Stable Video Diffusion use per query?
Each Stable Video Diffusion query consumes approximately 150.0 Wh of energy. This is 500x more than a traditional Google search (~0.3 Wh).
What is Stable Video Diffusion's carbon footprint?
Based on the carbon intensity of Self-hosted (varies), each query produces approximately 60.0 g of CO2. The grid in this region has a carbon intensity of 458 g CO2/kWh with 32% renewable energy.
How much water does Stable Video Diffusion use?
Each query consumes approximately 280 mL of water, primarily used for cooling the data centers that process the request.
How does Stable Video Diffusion compare to a Google search?
A Stable Video Diffusion query uses 500x more than a Google search in terms of energy. A Google search uses approximately 0.3 Wh, while Stable Video Diffusion uses 150.0 Wh.
Technical details
- Architecture
- Latent video diffusion model
- Parameters
- 1.5B
- Release date
- 2023-11-21
- Open source
- Yes