TBC’s Neuron Research Powers a Faster AI Video Model

The Biological Computing Co. (TBC) has released an image-to-video model that uses research involving living neurons to shape software. Called TBC mischief 1, the model is based on Wan 2.2 and became available on October 30 through fal, OpenRouter, and TBC’s creation app.

TBC says mischief 1 matches the quality of its base model while generating video five times faster. That comparison comes from the company, and the launch announcement does not include an independent benchmark or enough testing detail to show how the claimed gain transfers across different workloads.

Biology helps build the model, not run it

The name and research approach may suggest that customers send prompts to living cells or use biological processors. That is not how the system works. Mischief runs as software on conventional AI hardware, while TBC uses neuronal activity during its research and development process.

TBC’s website describes a sequence that starts by encoding computational challenges into electrical signals. The company then measures how cultured neuronal networks respond and extracts principles that can be implemented in software. The result is a digital inference system, not a biological computer that customers access when generating videos.

The company describes its commercial approach as a post-training software layer around diffusion transformer blocks. TBC says this layer adds less than 0.1% to the underlying model’s parameter count. In practical terms, the biological component belongs to discovery and development, while the model people use remains software running on conventional hardware.

That distinction is the central point behind TBC’s approach. The company is trying to shorten the loop for generating a short AI video by learning from measurements of living neurons, then translating those findings into a layer that modifies an existing model.

Prices, formats, and early production testing

OpenRouter lists mischief 1 at $0.005 per second for 480p video and $0.01 per second for 720p video. At the starting rate, a five-second clip costs $0.025, while 100 five-second clips cost $2.50.

The listing supports three- and five-second clips in a 16:9 aspect ratio, guided by a starting image. That makes the system an image-to-video tool rather than a text-only video generator. Users begin with an image, choose a supported clip length, and generate a short sequence from it.

Wonder Studios tested mischief 1 in its production workflow. In the launch announcement, TBC co-founder Justin Hackney describes faster, less expensive iteration as a benefit for creative work. The pricing and short clip formats fit that use case, where producing many versions can be part of the process.

Still, the five-times-faster figure needs to be read as a company-reported claim. The announcement does not establish the testing conditions, workloads, or independent results needed to determine whether the same improvement holds across different videos, resolutions, or production tasks.

The research behind TBC mischief 1

TBC’s Neural Dynamics Adapter technical post offers a more concrete example of the earlier research. In that work, the company recorded cultured neural activity using a high-density array with 4,096 electrodes. The recordings were converted into numerical representations, and the available source material ends mid-sentence after the words “and incorporat.”

That incomplete ending limits what can be stated about the technical post beyond the details provided. The broader process, however, is clear from TBC’s description: neuronal responses inform research, and the useful principles are translated into software rather than included as living components of the customer-facing system.

TBC mischief 1 therefore sits at the intersection of biological research and digital video generation. Its model is built on Wan 2.2, delivered through fal, OpenRouter, and TBC’s creation app, and priced by video duration on OpenRouter. The unusual part is not the hardware used for inference, but the way TBC says biological measurements helped shape the software layer around the existing model.

For now, the strongest performance statement remains TBC’s own claim that mischief 1 matches its base model’s quality while generating video five times faster. The release provides a concrete price, supported clip formats, and a description of the research path, but it does not provide the independent testing needed to confirm how broadly that speed claim applies.

Stay connected via Google News

Follow us for the latest AI News