What happened
The Biological Computing Co. (TBC) announced TBC Mischief 1 on October 30, 2026. The company describes it as a neuron-derived AI video model: it says that four years of studying living neurons grown on electrodes informed software efficiency techniques used in the release. Mischief 1 is built on Wan 2.2, an open-weight image-to-video model. TBC says its additions account for less than 0.1% of the base model's parameters and run on standard AI hardware.
The product is available as an API through two existing marketplaces, fal and OpenRouter. On fal, the listed image-to-video endpoint is tbc/mischief-1/image-to-video; on OpenRouter, it is tbc/mischief-1. The supplied listing information describes 480p and 720p output, 16:9 clips of either three or five seconds, and a first-frame image as the input. OpenRouter uses an asynchronous video workflow: an application submits a generation request, polls for completion, then downloads the result.
Pricing listed for OpenRouter starts at $0.005 per second for 480p output and $0.01 per second for 720p output. At the lower listed rate, a five-second clip costs $0.025. TBC calls Mischief 1 the lowest-priced image-to-video model on both marketplaces. That comparison is the company's positioning, rather than an independently supplied market survey, but the listed prices themselves give product teams a direct unit cost to test against their own workloads.
TBC also says Mischief 1 can generate video five times faster than the Wan 2.2 base model while matching its quality. The source does not provide an independent benchmark, prompts, sample set, latency measurements, or a methodology for that claim. The launch follows an earlier TBC announcement of an AWS collaboration for enterprise customers; the company says it is also working on language and physical-AI applications. This release, however, is specifically a public image-to-video offering through fal and OpenRouter.
What the archive adds
This is a new event in the broader hosted-video-model story, not a restatement of an earlier TBC announcement. Kling 4.0 video generation model available on Cloudflare AI Gateway and Workers AI documented another video model becoming accessible through an infrastructure gateway with published per-second pricing. Mischief 1 adds a different model and two different access surfaces, fal and OpenRouter.
Trend signal
The launch supports Generative media becomes application infrastructure. The important signal is not a standalone video demo: it is versioned access through common API marketplaces, bounded clip formats, and a visible usage price that an application team can put into a feature budget. The source does not establish whether this endpoint is reliable enough for production workloads.
Connections to other events
The earlier Kling 4.0 gateway release showed the same distribution pattern from a different direction: video generation is appearing behind developer-facing infrastructure with stated unit pricing. Mischief 1 differs by being offered through fal and OpenRouter and by positioning itself around a modified Wan 2.2 base model. Both events make it easier to compare video features as API choices rather than as isolated model announcements.
Why it matters
For teams building AI products, listed per-second pricing changes the first question from “can we afford a video feature?” to “which user action is worth a three- or five-second generation?” At $0.025 for a five-second 480p clip, teams can prototype use cases such as previsualization, animated product cards, creative variants, and high-volume internal experiments without beginning with a dedicated model-serving stack.
The availability on both fal and OpenRouter also reduces the amount of vendor-specific infrastructure needed to run an initial trial. That does not make model choice trivial. An application still needs to evaluate visual fidelity, motion consistency, latency, failure handling, rights and safety requirements, and how users react to short generated clips. Price is one operational variable, not a proxy for usable output quality.
What happens next
- More teams will run comparative image-to-video evaluations through hosted APIs rather than self-host Wan-derived models. (3 months, 65%). Watch for: published application integrations or benchmark comparisons using identical first-frame inputs.
- The listed cost advantage will be tested against output consistency and latency, not price alone. (3 months, 75%). Watch for: reproducible third-party evaluations of Mischief 1 against Wan 2.2 or other hosted video endpoints.
Limits and uncertainty
The public material is a company announcement. Its claims that Mischief 1 is five times faster than Wan 2.2, matches the base model's quality, and is the lowest-priced option on the two marketplaces are not accompanied here by independent testing, benchmark methodology, prompt sets, hardware details, latency distributions, availability guarantees, or quality comparisons. The material also does not establish commercial rights, rate limits, geographic availability, queue behavior, content-policy constraints, or whether the listed prices persist. Builders should treat the launch as an accessible endpoint and the performance narrative as a hypothesis to test.
Skill to build now
Build a small asynchronous video-evaluation harness this week. Start with a fixed set of representative first-frame images and prompts, submit three- and five-second jobs, store the returned output and elapsed completion time, and calculate cost from the listed per-second rates. Have reviewers score whether the generated motion preserves the intended subject and composition. This produces evidence for a product decision: which clip length, resolution, and workflow can meet a specific feature requirement, rather than which model has the strongest launch claim.
Reader takeaway
Mischief 1 is a new hosted image-to-video option with publicly listed prices from $0.005 per second and access through fal and OpenRouter. That makes it straightforward to trial in a product workflow. The claimed speed and quality advantage over Wan 2.2 has not been independently established in the supplied material, so evaluate it on representative jobs before designing around it.
Open question
Will independent tests show that Mischief 1's lower advertised cost and claimed speed retain acceptable motion quality and completion latency for production-facing video features?