Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Response to Amendment
The Amendment filed October 22nd, 2025 has been entered. Claims 1, 9-12, 16 and 18-20 have been amended. Claims 1-20 remain rejected in the application. Applicant’s amendments to the drawings, specifications and claims have overcome each and every objection previously set forth in the Non-Final Office Action mailed July 23rd, 2025 and have therefore been withdrawn.
Claim Rejections - 35 USC § 102
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
(a)(2) the claimed invention was described in a patent issued under section 151, or in an application for patent published or deemed published under section 122(b), in which the patent or application, as the case may be, names another inventor and was effectively filed before the effective filing date of the claimed invention.
Claims 1-6, 10-15 and 19-20 are rejected under 35 U.S.C. 102(a)(2) as being anticipated by Bradley et al. (Pub. No.: US 2023/0154090 A1), hereinafter Bradley.
Regarding claim 1, Bradley discloses a computer-implemented method for rendering an image of a face (FIG. 4 and paragraph 52 teach that FIG. 4 is a flow diagram of method steps for synthesizing a sequence of 3D geometries, according to various embodiments), the method comprising:
performing one or more operations to convert a texture map associated with the face to a neural texture map (Paragraph 55 teaches that in step 406, execution engine 124 converts, via the encoder neural network, the input geometries into one or more latent vectors and paragraph 60 teaches that in one or more embodiments, generator 500 includes components that generate neural textures 538, given input vectors 536 that are sampled from one or more distributions. In some embodiments, neural textures 538 include representations of textures that are generated by one or more neural network layers for one or more portions of a 3D geometry (e.g., geometries 218). These neural textures 538 are combined with one or more texture maps 532 and/or one or more segmentation masks 534 that are generated from the 3D geometry to form an image (e.g., images 540) that corresponds to a rendering of the 3D geometry. Additionally, FIG. 6C and Paragraphs 68 through 71 teach that FIG. 6C illustrates a number of maps 622, 624, 626, 628, and 630 that are used to sample and composite neural textures 538 from the exemplar generator 500 of FIG. 6A, according to various embodiments. As shown in FIG. 6C, map 622 includes a texture map of the skin, eyes, and inner mouth in a face geometry, and map 624 includes a texture map of the hair in the face geometry. Maps 622 and 624 can be generated by posing and rendering components 612, 614, 616, and 618 of a deformed face model, as described above with respect to FIG. 6B. Map 626 includes a segmentation mask of the face geometry, and maps 628 include intermediate neural textures 538 for various components of the face geometry. Map 626 can also be generated by rendering the deformed face model in a certain pose, and maps 628 can be generated by individual generator blocks 602 in generator 500. Finally, map 630 includes composited screen-space neural features 604 for the face geometry. Map 630 can be generated by sampling neural textures 538 in maps 628 using the corresponding texture maps 622 and 624 and assembling and layering the sampled neural textures 538 using the segmentation mask in map 626. While the operation of generator 500 has been discussed with respect to FIGS. 6A-6C in the context of face geometries and face models, those skilled in the art will appreciate that generator 500 can be used to perform rendering of other types of objects and/or geometries. For example, generator blocks 602 could be used to generate neural textures 538 for various body parts of a human or animal. These neural textures 538 could be combined with texture maps 532 for the same body parts to generate screen-space neural features for each of the body parts. A segmentation mask of the body parts could then be used to composite the screen-space neural features, and one or more convolutional layers 606 in generator 500 could be used to convert the composited screen-space neural features 604 into a rendered image 608 of the human or animal.);
and performing, via a first trained machine learning model, one or more operations to generate the image of the face based on the neural texture map and first 3D geometry associated with the face (Paragraph 57 teaches that in step 410, execution engine 124 causes output related to the animation to be generated based on the sequence of geometries. For example, execution engine 124 could store the sequence of geometries and/or corresponding input geometries in a training dataset for the encoder neural network, decoder neural network, and/or another machine learning model. In another example, execution engine 124 could transmit the sequence of geometries to an application or service that generates animations and/or other types of graphical or geometric output based on the sequence of geometries. Additionally, paragraph 67 teaches that in one or more embodiments, components 612, 614, 616, and 618 are assembled within the face model and rendered to produce corresponding texture maps 532 that are used to sample UV-space neural textures 538. More specifically, a template for the face model can be deformed to match the identity and expression of an input face geometry. The deformed face model is then posed and rendered to produce texture maps 532 and a segmentation mask for the input face geometry).
Regarding claim 2, Bradley discloses everything claimed as applied above (see claim 1), in addition, Bradley discloses performing one or more operations to convert a lighting map associated with the face to a neural lighting map, wherein the image of the face is further generated based on the neural lighting map (Paragraph 85 teaches that the technique of FIG. 7 can additionally be used to control other aspects of a rendered image. For example, latent vector 710 could be divided into components that represent lighting, pose, age, background, accessories, proportions, and/or other attributes related to the appearance of a face (or another object) in an image produced by the generative model. Training data that includes images of the same identities, variations in these attributes, and distinct coefficients or values that represent these variations in attributes could be used to train the generative model. The trained generative model could then be used to generate images of specific identities and/or attributes).
Regarding claim 3, Bradley discloses everything claimed as applied above (see claim 1), in addition, Bradley discloses wherein the first trained machine learning model comprises an encoder that encodes the neural texture map and the first 3D geometry to an embedding, and a decoder that decodes the embedding to generate the image of the face (FIG. 2, FIG. 3 and paragraph 33 teach that as shown in FIG. 2, transformer 200 includes an encoder 204 and a decoder 206. In various embodiments, encoder 204 and decoder 206 are implemented as neural networks. Additionally, paragraph 39 teaches that after latent vectors 222 are generated as the output of the last encoder block 306(Y) in encoder 204, decoder 206 is used to generate a full synthesized sequence 216 of geometries that includes input geometries 220. As shown in FIG. 3, input 312 into decoder 206 includes a position-encoded capture code 224. As mentioned above, capture code 224 encodes the content, speed, context, semantics, identity, and/or other aspects of synthesized sequence 216. For example, capture code 224 includes a “d-dimensional” vector that represents an actor, speaking style, speed, semantics, or other attributes of a facial or full-body performance from which synthesized sequence 216 is to be generated. In various embodiments, this vector is obtained as an embedding from one or more layers of encoder 204 and/or decoder 206 and/or from an external source).
Regarding claim 4, Bradley discloses disclose everything claimed as applied above (see claim 3), in addition, Bradley discloses disclose performing one or more operations to train the encoder and one or more layers of the decoder while keeping one or more pre-trained layers of the decoder fixed (Paragraph 72 teaches that returning to the discussion of FIG. 5, training engine 132 trains generator 500 using generator training data 514 that includes training texture maps 528 and training segmentation masks 530 associated with a number of synthetic geometries 526. Synthetic geometries 526 include 3D models of synthetic objects that are similar to objects for which images 540 are to be generated. For example, synthetic geometries 526 could include full-head 3D models of synthetic faces. Training engine 132 and/or another component could generate each synthetic face by randomizing the identity, expression, hairstyle, and/or pose of a parametric face model, such as the face model of FIG. 6B. The component could then generate one or more training texture maps 528 and/or one or more training segmentation masks 530 for each synthetic face by posing and rendering the corresponding face model, as described above with respect to FIGS. 6B-6C. Additionally, paragraph 78 teaches that consequently, execution engine 134 can use generator 500 to produce images 540 of fixed geometries 218 and/or neural textures 538. More specifically, execution engine 134 can keep input vectors 536 fixed to generate the same neural textures 538 across multiple images 540).
Regarding claim 5, Bradley discloses everything claimed as applied above (see claim 1), in addition, Bradley discloses inputting the texture map into a second trained machine learning model that outputs the neural texture map (Paragraph 90 teaches that as shown, in step 902, training engine 132 trains one or more neural networks based on a training dataset that includes texture maps, segmentation masks, and/or styles for a set of synthetic geometries. For example, training engine 132 could train a generator neural network and/or an image-to-image translation network to generate RGB images of each synthetic geometry, given the corresponding texture maps, segmentation masks, and/or a set of blendshape coefficients representing an “expression” style associated with the synthetic geometry. Training engine 132 could also, or instead, train the generator neural network and/or image-to-image translation network in an adversarial fashion based on predictions generated by a discriminator neural network from images produced by the generator neural network and/or the image-to-image translation network).
Regarding claim 6, Bradley discloses everything claimed as applied above (see claim 5), in addition, Bradley discloses wherein the second trained machine learning model comprises a convolutional neural network (Paragraph 87 teaches that the segmentation mask 802 is inputted into a convolutional neural network (CNN) 804 that performs image-to-image translation).
Regarding claim 10, Bradley discloses everything claimed as applied above (see claim 1), in addition, Bradley discloses performing one or more training operations to generate the first trained machine learning model based on one or more other images that include the face (Paragraph 72 teaches that training engine 132 and/or another component could generate each synthetic face by randomizing the identity, expression, hairstyle, and/or pose of a parametric face model, such as the face model of FIG. 6B. The component could then generate one or more training texture maps 528 and/or one or more training segmentation masks 530 for each synthetic face by posing and rendering the corresponding face model, as described above with respect to FIGS. 6B-6C).
Regarding claim 11, Bradley discloses everything claimed as applied above (see claim 1), in addition, Bradley discloses performing one or more training operations to generate the first trained machine learning model based on a plurality of images associated with a plurality of different faces (Paragraph 74 teaches that in one or more embodiments, training engine 132 updates parameters of generator 500 based on predictions 512 outputted by a discriminator 510 from training images 508. As shown in FIG. 5, input into discriminator 510 includes training images 508 produced by generator 500 from generator training data 514, as well as images 522 from discriminator training data 516 for discriminator 510. For example, training images 508 could include images of faces that are rendered by generator 500 using training textures 502, samples 504, and composited features 506, and images 522 could include photographs of faces).
Regarding claim 12, the non-transitory computer-readable media steps correspond to and are rejected the same as the method steps of claim 1 (see claim 1 above), in addition, Bradley discloses a non-transitory computer-readable media storing instructions and at least one processor (Paragraph 112 teaches that in some embodiments, one or more non-transitory computer readable media store instructions that, when executed by one or more processors, cause the one or more processors to perform the steps of generating one or more maps associated with one or more portions of a first input geometry).
Regarding claim 13, the non-transitory computer-readable media steps correspond to and are rejected the same as the method steps of claim 2 (see claim 2 above).
Regarding claim 14, the non-transitory computer-readable media steps correspond to and are rejected the same as the method steps of claim 3 (see claim 3 above).
Regarding claim 15, the non-transitory computer-readable media steps correspond to and are rejected the same as the method steps of claim 5 (see claim 5 above).
Regarding claim 19, the non-transitory computer-readable media steps correspond to and are rejected the same as the method steps of claim 10 (see claim 10 above).
Regarding claim 20, the system steps correspond to and are rejected the same as the method steps of claim 1 (see claim 1 above), in addition, Bradley discloses a system (FIG. 1 and paragraph 22 teach that FIG. 1 illustrates a computing device 100 configured to implement one or more aspects of various embodiments), comprising:
a memory storing instructions (Paragraph 22 teaches that the computing device 100 is configured to run a geometry synthesis module 118 and an image synthesis module 120 that reside in a memory 116);
and a processor that is coupled to the memory (Paragraph 24 teaches that in one embodiment, computing device 100 includes, without limitation, an interconnect (bus) 112 that connects one or more processors 102, an input/output (I/O) device interface 104 coupled to one or more input/output (I/O) devices 108, memory 116, a storage 114, and a network interface 106).
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 7 and 16 are rejected under 35 U.S.C. 103 as being unpatentable over Bradley in view of Chae et al. (Pub. No.: US 2022/0351348 A1), hereinafter Chae.
Regarding claim 7, Bradley discloses everything claimed as applied above (see claim 5), however, Bradley fails to disclose performing one or more operations to train a first machine learning model and a second machine learning model simultaneously to generate the first trained machine learning model and the second trained machine learning model, respectively.
Chae discloses performing one or more operations to train a first machine learning model and a second machine learning model simultaneously to generate the first trained machine learning model and the second trained machine learning model, respectively (FIG.3 and paragraph 80 teach that referring to FIG. 3, a speech moving image generation device 200 may include a first machine learning model 202, a second machine learning model 204, and a third machine learning model 206. The first machine learning model 202, the second machine learning model 204, and the third machine learning model 206 may form one neural network model through which they are interconnected. That is, the first machine learning model 202, the second machine learning model 204, and the third machine learning model 206 may be organically connected and simultaneously trained). Since Bradley teaches the initial method steps for converting and training one or more machine learning models to output neural texture maps and Chae teaches training multiple machine learning models at the same time, simultaneously, it would have been obvious to a person having ordinary skill in the art to combine the features together so that while training the different machine learning models, they could each be trained at the same time, which would help reduce the overall time that it would take to train multiple machine learning models.
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Bradley to incorporate the teachings of Chae, so that the combined features together would improve and reduce the overall amount of time needed to train the different machine learning models, by training them simultaneously.
Regarding claim 16, the non-transitory computer-readable media steps correspond to and are rejected the same as the method steps of claim 7 (see claim 7 above).
Claims 8-9 and 17-18 are rejected under 35 U.S.C. 103 as being unpatentable over Bradley in view of Chae (Pub. No.: US 2022/0375224 A1).
Regarding claim 8, Bradley discloses everything claimed as applied above (see claim 1), in addition, Bradley discloses generating a second 3D geometry and the texture map based on a face associated with an actor included in a first video frame of a first media content item (Paragraph 91 teaches that in step 904, execution engine 134 generates a segmentation mask and/or one or more texture maps associated with one or more portions of an input geometry and paragraph 106 teaches further comprising rendering a second image corresponding to a second input geometry based on a second segmentation mask for the second input geometry, a second set of texture maps for one or more portions of the second input geometry, and the first set of neural textures. Additionally, paragraph 35 teaches that in some embodiments, input 302 into encoder 204 includes position-encoded representations of a number of input geometries 220. In various embodiments, these position-encoded representations are generated by combining input geometries 220 with position encodings 304 that represent the positions of the corresponding frames within the animation. For example, input 302 could be generated by adding a “positional encoding” that represents the position (e.g., frame number, time step, etc.) of each input geometry within a performance to a mesh, a set of blendshape weights, an embedding, and/or another representation of the input geometry). However, Bradley fails to disclose generating a third 3D geometry based on a face associated with a dubber included in a second video frame of a second media content item.
Chae discloses generating a third 3D geometry based on a face associated with a dubber included in a second video frame of a second media content item (Paragraph 88 teaches that the first encoder 102 may extract an image feature vector from the person background image, paragraph 89 teaches that the second encoder 104 receives an input of a speech audio signal. Here, the speech audio signal may not be related to the person background image input to the first encoder 102. For example, the speech audio signal may be a speech audio signal of a person different from the person in the person background image and paragraph 90 teaches that the combining unit 106 may generate a combined vector by combining the image feature vector output from the first encoder 102 and the voice feature vector output from the second encoder 104). Since Bradley teaches generating different texture maps for an initial face of a person/actor depending on certain frames and Chae teaches generating image feature vectors and voice feature vectors related to another person’s/dubber’s speech or audio, it would have been obvious to a person having ordinary skill in the art to combine the features together so that in addition to being able to generate a texture map related a first person/actor and a particular frame, the texture map and audio features of a second person could also be generated as well.
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Bradley to incorporate the teachings of Chae, so that the combined features together would allow for additional 3D geometries to be generated, including those related to a second person.
Additionally, Bradley in view of Chae disclose and performing one or more operations to align the third 3D geometry with the second 3D geometry to generate the first 3D geometry (Paragraph 56 if Chae teaches that the person background image input to the first encoder 102 and the speech audio signal input to the second encoder 104 may be synchronized in time. Additionally, paragraph 92 of Chae teaches that the second decoder 110 may predict and output a landmark of the speech video using the combined vector as an input. Here, since the speech video generation device 100 is trained to also predict a landmark of a speech video while reconstructing the speech video through the first decoder 108 and the second decoder 110 when the combined vector is input, the speech video generation device 100 may accurately and smoothly predict without a process of aligning the landmark of the speech video in a standard space).
Regarding claim 9, Bradley discloses everything claimed as applied above (see claim 1), in addition, Bradley discloses generating a second 3D geometry and the texture map based on a face associated with an actor included in a first video frame of a first media content item (Paragraph 106 teaches further comprising rendering a second image corresponding to a second input geometry based on a second segmentation mask for the second input geometry, a second set of texture maps for one or more portions of the second input geometry, and the first set of neural textures). However, Bradley fails to disclose generating a third 3D geometry associated with another face based on audio associated with a dubber included in a second media content item.
Chae discloses generating a third 3D geometry associated with another face based on audio associated with a dubber included in a second media content item (Paragraph 57 teaches that the combining unit 106 may generate a combined vector by combining an image feature vector output from the first encoder 102 and a voice feature vector output from the second encoder 104. In an example embodiment, the combining unit 106 may generate the combined vector by concatenating the image feature vector and the voice feature vector, but the present disclosure is not limited thereto, and the combining unit 106 may generate the combined vector by combining the image feature vector and the voice feature vector in other various manners. Additionally, paragraph 55 teaches that the second encoder 104 may be a machine learning model trained to extract a voice feature vector using a speech audio signal as an input. Here, the speech audio signal corresponds to an audio part in the person background image (i.e., an image in which a person utters) input to the first encoder 102. In other words, a video part in a video in which a person utters may be input to the first encoder 102, and an audio part may be input to the second encoder 104). Since Bradley teaches generating different texture maps for an initial face of a person/actor depending on certain frames and Chae teaches generating different facial image features based on audio portions related to that person/actor, it would have been obvious to a person having ordinary skill in the art to combine the features together so that in addition to being able to generate a texture map related a first person/actor and a particular frame, the audio features associated with that person’s face or another person’s face, could also be used as well.
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Bradley to incorporate the teachings of Chae, so that the combined features together would allow for additional 3D geometries to be generated, including those related to another person speaking or their audio.
Additionally, Bradley in view of Chae disclose and performing one or more operations to align the third 3D geometry with the second 3D geometry to generate the first 3D geometry (Paragraph 64 of Chae teaches that the time of the person background image input to the first encoder 102 and the time of the speech audio signal input to the second encoder 104 may be synchronized with each other).
Regarding claim 17, the non-transitory computer-readable media steps correspond to and are rejected the same as the method steps of claim 8 (see claim 8 above).
Regarding claim 18, the non-transitory computer-readable media steps correspond to and are rejected the same as the method steps of claim 9 (see claim 9 above).
Response to Arguments
Applicant's arguments filed October 22nd, 2025 have been fully considered but they are not persuasive. In response to applicant's argument that the amendments to claim 1 appear to show that the prior art of Bradley contains no such teachings of one or more texture maps being converted into one or more neural textures, a recitation of the intended use of the claimed invention must result in a structural difference between the claimed invention and the prior art in order to patentably distinguish the claimed invention from the prior art. If the prior art structure is capable of performing the intended use, then it meets the claim.
PNG
media_image1.png
216
429
media_image1.png
Greyscale
According to FIG. 6C and paragraph 68 of Bradley, FIG. 6C illustrates a number of maps 622, 624, 626, 628, and 630 that are used to sample and composite neural textures 538 from the exemplar generator 500 of FIG. 6A, according to various embodiments. Additionally, paragraph 69 goes on to teach that map 626 includes a segmentation mask of the face geometry, and maps 628 include intermediate neural textures 538 for various components of the face geometry. Lastly, paragraph 70 teaches that map 630 includes composited screen-space neural features 604 for the face geometry. Map 630 can be generated by sampling neural textures 538 in maps 628 using the corresponding texture maps 622 and 624 and assembling and layering the sampled neural textures 538 using the segmentation mask in map 626.
In appears that the prior art of Bradley does have a structure that is capable of performing a similar functionality to the claimed step of performing one or more operations to convert a texture map associated with the face to a neural texture map due to the fact that the different maps (especially map 628 that includes different sampled neural textures 538) when combined with each other can generate map 630, which is a map generated by sampling the different neural textures 538 and assembling and layering them together with the other maps to produce (and thus convert them into) a newly generated neural texture map. This suggests that the prior art of Bradley is indeed able to perform the intended use present since, the combining of the different texture maps, changes and converts them, into a type of neural texture map.
In response to the argument pertaining to the interpretation of the claim language of “texture map”, the argument is considered moot since these additional citations related to FIG. 6C point out that the different maps used for generating the neural texture map, do include different types of texture maps (622 and 624).
In regards to similar arguments related to the amended independent claims of 12 and 20, they are rejected similarly to the reasoning for claim 1. In addition, any arguments regarding any of the dependent claims 2-11 and 13-19, for the virtue of their dependency are moot because the independent claims are not allowable.
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action. 12. Any inquiry concerning this communication or earlier communications from the examiner should be directed to George Renze whose telephone number is (703)756-5811. The examiner can normally be reached Monday-Friday 9:00am - 6:00pm EST.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Xiao Wu can be reached at (571) 272-7761. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/G.R./Examiner, Art Unit 2613
/XIAO M WU/Supervisory Patent Examiner, Art Unit 2613