The Biological Computing Co. launched TBC mischief 1 on October 30, 2026, as a Wan 2.2-based image-to-video model available through fal and OpenRouter. The company attributes its optimization approach to neuron-derived research. Hosted video generation runs as software; the announcement does not establish live-neuron inference for each request.
What became available on October 30
OpenRouter dates the release to October 30 and lists the model as tbc/mischief-1. fal supplies both a playground and API documentation for tbc/mischief-1/image-to-video. That establishes a publicly listed hosted service, rather than merely a preview video or a future access promise. It does not establish that every account, region or workload will be supported.
The distinctions matter. An announcement describes the product; a playground lets users submit examples; API availability enables application integration. Downloadable weights would enable a different deployment path. The reviewed launch and marketplace pages establish hosted access, but do not identify a mischief 1 weights download. Wan 2.2’s open-weight status should not be mistaken for a downloadable release of TBC’s derivative.
The neurons inform the software; they do not serve each request
TBC’s earlier September 23 AWS collaboration announcement explains the delivery model: measurements of living neural activity inform a proprietary software layer running on conventional AI infrastructure. Customers do not need biological hardware. That announcement described early access and plans for Trainium, SageMaker AI and AWS Marketplace distribution; it was not proof that those deployment routes were already available.
The October 30 launch names fal and OpenRouter as live access channels. For procurement, the practical interpretation is ordinary hosted-model evaluation, not buying or maintaining a biological computer. The reviewed evidence does not independently establish how much of the reported efficiency gain comes specifically from biological measurements.
Short clips, resolution-specific prices and workflow constraints
fal’s schema requires a starting image and a motion prompt. It documents 480p or 720p landscape output, three- or five-second durations, and MP4 video at 16 frames per second. This is a short image-animation endpoint, not evidence of a released long-form, text-only video system.
At the November 1 review, fal lists $0.005 per generated second at 480p and $0.01 at 720p. Its listed five-second examples therefore cost $0.025 and $0.05 respectively. The lower headline rate should not be applied to both resolutions. Those are published generation charges, not a measured cost per approved creative asset.
There are operational details worth preserving in a pilot. fal says a fixed seed alone does not control prompt rewriting; reproducible rendering requires original prompt processing. Its queue interface supports submission, status checks and result retrieval. OpenRouter also flags that the model is not currently available through EU in-region routing, a limitation buyers with regional-processing requirements should resolve before uploading assets.
Why the next useful result is an acceptance test
Our analysis: the strongest initial business case is a bounded creative experiment. Ask whether animating approved stills accelerates concept review—not whether an unusual research origin guarantees better finished advertising. A sensible pilot would compare:
- Time from submission to a downloaded, reviewable clip, including retries.
- Accepted clips per batch, judged against the same product-shape and motion requirements.
- Total generation spend divided by accepted clips, plus human review and editing time.
Keep images, requested duration, resolution and acceptance criteria consistent. Log rejected outputs rather than counting every successful request as useful work. These are proposed evaluation measures, not results from firsthand testing.
The October 30 release warrants attention because access can now be evaluated. Whether the claimed acceleration produces business value still depends on usable outputs, dependable delivery and deployment terms—not the speed headline alone.