The Biological Computing Co. (TBC) released Mischief 1 on 30 October and calls it the first publicly available “neuron-derived” AI video model. Strip the framing and it is an image-to-video model you can call today on OpenRouter and fal.
What it is
Mischief 1 is built on Wan 2.2, the open video model from Alibaba’s Wan team. TBC adds an adapter that it says is less than 0.1% of the model’s parameters. The company says it designed that adapter using signals recorded from living neurons grown in its lab. The neurons stay in the lab: what ships is ordinary model weights running on ordinary GPUs, so “runs on brain cells” would be wrong. TBC claims the result generates video 5 times faster than the base model. It has not published a benchmark anyone else can rerun, so treat that number as the company’s.
Price and limits
OpenRouter lists Mischief 1 from $0.005 per second at 480p and $0.01 per second at 720p, released 30 October. A 5-second 480p clip is 5 times $0.005, or 2.5 cents. The listing says it makes 3- and 5-second clips in 16:9 and is steered by a first-frame image; the fal page lists the image-to-video endpoint with commercial use allowed.
For scale, our note on Grok Imagine Video 1.5 Lite covers a pricier option that does longer clips with speech. If you are new to how these models turn a still into motion, our explainer on AI video generation covers the basics.
Why this matters
Half a cent a second makes Mischief 1 one of the cheapest hosted video endpoints we have priced, and it sits behind APIs many teams already use, so testing it costs very little. The neuron story is interesting research, but there is no independent evidence yet that it beats a well-tuned Wan 2.2 on quality. Run your own prompts against the base model before choosing it for anything beyond short drafts.
The Campfire
CommentsThanks. Your comment is waiting for the desk. Once it is approved it appears here within about a minute.
Add a comment