Federal Judge Invalidates Four Text-to-Video Patents as Abstract Ideas, Ending Suit Against OpenAI

SAN FRANCISCO (LN) — A federal judge dismissed with prejudice a patent infringement suit by intellectual property holding company MITII, Inc. against OpenAI OpCo, LLC, ruling that MITII's four text-to-video patents claim an abstract idea and offer no inventive concept sufficient to survive scrutiny under the two-step Alice framework.

MITII sued OpenAI in October 2025, alleging that Sora, OpenAI's text-to-video app, infringed four patents — U.S. Patent Nos. 9,397,972, 9,667,574, 10,616,157, and 11,005,796 — directed to technology that converts a typed message into a short animated video of a well-known character speaking the text. The patents were originally described as helping children with autism by incorporating familiar characters into text-to-video media.

U.S. District Judge Noël Wise treated claim 1 of the '972 patent as representative of all four, finding MITII had failed to make meaningful arguments about the distinctive significance of the remaining claims. The representative claim lays out 14 steps — composing a message, selecting an animation character, transmitting both to a server, converting the message to speech, generating moving images, and displaying the result on a second device — but, Judge Wise found, never explains how any of those conversions actually work.

"Claim 1 does not specify how an electronic message is converted into speech or moving images of a well-known animation character are generated," Judge Wise wrote, concluding the claim "simply add[s] conventional computer components" to well-known practices.

At Alice step one, Judge Wise rejected MITII's argument that the claims were analogous to those upheld in McRO, Inc. v. Bandai Namco Games Am. Inc., where the Federal Circuit upheld claims tied to rules with specific characteristics governing morph weight sets. The '972 patent's claim 1, she held, is not focused on a specific means or method that improves animation but rather is directed to a result — the generation of a talking animated character from a text message — without disclosing how to achieve it.

At step two, MITII argued the use of well-known characters' likenesses was unconventional because unauthorized use carries civil liability, and that video generation using a famous character's likeness was not routine. Judge Wise was unpersuaded on both counts, noting that the civil-liability argument was "untethered from representative claim 1," which says nothing about licensed use or liability, and that MITII had failed to explain why the concept was non-routine when the claim itself relies only on generic servers and electronic devices.

MITII also sought leave to amend, but Judge Wise denied it, citing the Federal Circuit's holding that no amendment to a complaint can alter what a patent itself states.

The case was originally filed in the Central District of California stipulated to transfer it to the Northern District in early January 2026.

MITII did not respond to a request for comment. OpenAI had no immediate statement.

The patents at issue employ text-to-speech technology and computer animation techniques that, Judge Wise found, existed of filing — leaving MITII with no path to replead its way back into court.