McRO, Inc., owns a U.S. patent on a method for animating faces in video games. Back in 2016, when McRO sued a number of video-game developers for patent infringement, the U.S. Court of Appeals for the Federal Circuit (CAFC) upheld the patent as subject-matter eligible. Now, the CAFC has again upheld the patent, this time against a challenge of non-enablement.
Patents must enable the full scope of claims
At their core, patents are a quid pro quo bargain. In return for patent rights, the inventor must teach the public how to practice the invention. The standard is: “After reading this disclosure, would a person skilled in the art know — without undue experimentation — how to practice the invention?” This is stated in Section 112 as the enablement requirement. If the written disclosure does not fully enable the claimed invention, then the inventor has not held up their end of the bargain and the patent is invalid. One ready way of attacking a patent as non-enabling is to point to an embodiment of the claimed invention that would not be understood from the disclosure. For example, if a patent claims pens of all colors, but only teaches how to make blue pens, then the patent is non-enabling and therefore invalid.
To show non-enablement, the defendants pointed to two methods of animation that were not fully enabled by the patent’s specification. But when the court found that neither method was infringing — and thus that both methods were outside the scope of the claims — the defendants offered no other concrete examples. They were left with an “abstract assertion” that the claims covered non-enabled methods. The CAFC held that an “abstract assertion of breadth” will not support a claim of non-enablement.
Animating speech in video games
When video-game characters speak, they sometimes look like sock puppets — the mouth opens and closes, the head might turn or nod. To realistically animate speech, designers must match the mouth shape to the character’s words and the facial expression to their emotions. Manually drawing that, frame by frame, takes too long. McRO’s patent covers a method for automatically generating the lip movements and facial expressions of speaking characters.
The method works by first associating each possible single sound (a phoneme) with one mouth position (a morph target). The patent describes specifying the mouth model by identifying vertices in particular positions. Then each morph target is defined as the set of vectors which describe the displacement of each vertex from its position on a neutral, reference mouth. Each facial expression is described by a blending of multiple morph targets in specific proportions, known as a “morph weight set.” This much of the method was known in the prior art.
The novelty of McRO’s patent is in automatically generating a stream of morph weight sets from a time-marked transcript of recorded phonemes. Before, artists would have had to assign the morph weight sets at each frame. But McRO’s patent describes using a set of rules to automatically generate the stream of morph weight sets, given only a time-aligned phonetic transcription.
The ’278 Patent
At issue were three method claims of U.S. Patent No. 6,611,278, of which claim 1 is representative:
- A method for automatically animating lip synchronization and facial expression of three-dimensional characters comprising:
obtaining a first set of rules that defines a morph weight set stream as a function of phoneme sequence and times associated with said phoneme sequence;
obtaining a plurality of sub-sequences of times phonemes corresponding to a desired audio sequence for said three-dimensional characters;
generating an output morph weight set stream by applying said first set of rules to each sub-sequence of said plurality of sub-sequences of timed phonemes; and
applying said output morph weight set stream to an input sequence of animated characters to generate an output sequence of animated characters with lip and facial expression synchronized to said audio sequence.
Under the CAFC’s construction in 2016, in the field of rules-based animation, the claims of the ’278 Patent are limited by defining morph weight sets as a function of the timing of phoneme subsequences. In the specification, morph weight set is defined in terms of a “delta set,” i.e., the set of displacement “vectors” of each vertex from the reference mouth. Critically, the CAFC agreed with the district court that these vectors must be three-dimensional.
Bones and BALDI
The defendant developers use “bones animation” to animate faces. As in the ’278 Patent, in bones animation, models are specified by identifying vertices and transformations of their positions. But in bones animation, the transformations are encoded as a four-by-four matrix defining movements which can include not only translations, but also rotations and scales. This is in contrast to the three-dimensional displacement vectors to which the ’278 Patent was limited by the court. Consequently, the district court had held that developers’ use of bones animation did not infringe McRO’s patent.
A second animation method the court calls “the BALDI system.” According to the district court, the BALDI system uses an interlocking set of mathematical equations that balance the influence of different proximate phonemes on various aspects of the model’s facial expression. As with bones animation, BALDI’s equations do not represent displacements of vertices by simple three-dimensional vectors.
Both bones and BALDI animation, then, do not infringe the claims of the ’278 Patent. The district court so held, and the CAFC agreed.
Abstract assertion of breadth
In addition to its finding of non-infringement, the district court had also invalidated the ’278 Patent for non-enablement. Section 112 requires that a patent enable a person skilled in the art to practice the full scope of the claimed invention. Importantly, a patent need only enable the claimed invention, not matter outside the claims. So the question of enablement depends on an interpretation of the claims — what they cover and what they don’t.
Following this logic, a finding of non-enablement requires “concrete identification of at least some embodiment or embodiments asserted not to be enabled.” In re Wands, 858 F.2d 731, 737 (Fed. Cir. 1988). The CAFC cited a number of cases to emphasize that the identification must be concrete, not abstract. To prove non-enablement, it is not enough to abstractly explain that that the claims might cover more than the disclosure enables. Instead, the courts will require a concrete example.
The developers pointed to two such concrete examples — bones animation and BALDI animation. They claimed that neither method was enabled by the specification. For example, each method uses transformations of more dimensions that the 3D vectors disclosed by the ’278 Patent. The developers attempted to show that the two methods were within the scope of the claims but were not enabled. As such, the ’278 Patent would not have enabled the full scope of the claimed invention and would therefore be invalid.
However, as discussed above, the developers had also argued that these methods were non-infringing — and the court agreed! The court interpreted the claims as reading on vectors only of three dimensions, leaving bones and BALDI outside the claims’ scope. This also removed them from the scope of methods which the ’278 Patent could be required to enable.
Without the bones and BALDI examples, the developers were left only with an “abstract assertion” that the claims exceeded the enabled scope. In particular, they focused on the “first set of rules” limitation in claim 1. This is conceivably broad, and the district court had found that it was broader than anything disclosed in the specification. The CAFC disagreed and narrowly construed the limitation according to the specification. It went even further and found that the “first set of rules” limitation would have been well known in the art, whether the ’278 Patent disclosed it or not. Finally, without any concrete example of a non-enabled embodiment, the developers’ “abstract assertion of [the] breadth” of the claims could not carry their burden under the Wands.
This case is a good example of the interplay between enablement and claim interpretation. To win on non-infringement, the developers had argued for a narrow reading of the claims. But the narrow claims were fully enabled. Had they lost on infringement, they might have won on enablement. An interesting tension in itself, patent litigators can learn an important lesson from this case. Like Hercules at the crossroads, choose carefully.
Be careful how you choose your arguments. Lawyers are accustomed to arguing in the alternative, offering a number of arguments — even taking inconsistent positions — in the hope that one will carry the day. But if you aren’t careful, a win could lead to a loss. Certainly, the defendants were happy to get a judgment of non-infringement, so that they can keep doing what they do. But they likely would also have like a judgment of invalidity, so that they wouldn’t have to worry about the ’278 Patent in the future.
This might have been avoided with surgically precise claim constructions. The developers did not actually practice the BALDI method, only the bones method. Had they been able to get a claim interpretation that read on BALDI but not bones, the developers could have used BALDI as a concrete example for meeting their burden under Wands. They’d have gotten two victories for the price of one.