DETAILED ACTION
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 .
Continued Examination Under 37 CFR 1.114
A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant's submission filed on 12/29/2025 has been entered.
Response to Arguments
Applicant's arguments filed 12/29/2025 have been fully considered but they are not persuasive.
With respect to applicant’s argument that the Yu reference fails to disclose a multilingual text to image generator because paragraph 0020 of Yu instead discloses a monolingual text to image generator that is aligned to a multilingual pre-trained language model, the examiner maintains. The Yu reference discloses utilization of a multilingual text to image generator (paras 0030, para 0101: multilingual M-T21 model).
Claim Rejections - 35 USC § 102
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)(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 1-2, 6-7, 11-16 & 18-19 are rejected under 35 U.S.C. 102(a)(2) as being anticipated by Yu et al, US Patent Pub. 20240185035 A1.
Re Claim 1, Yu et al discloses a method comprising: obtaining a text prompt in a first language (abstract: text-to-image model receives text prompts and generates images); encoding the text prompt using a multilingual encoder to obtain a multilingual text embedding (paras 0030, 0101: multilingual model implies multilingual (multiple languages) text embedding capabilities utilizing the text prompts to generate images via the text-to-image diffusion models as highlight in (abstract); wherein the process is carried out via processors with accompanying memory (para 0040)); processing the multilingual text embedding using a diffusion prior model to obtain an image embedding (paras 0076-0085: image generator includes various steps including an image feature embedding being generated before generating a noise image embedding before generating a noise reduced image embedding before ultimately generating the final output image from the text prompt), wherein the diffusion prior model is trained to perform denoising based on multilingual text embeddings from the first language and a second language to obtain input guidance for an image generation model different from the diffusion prior model (paras 0076-0085: image generator includes various steps including an image feature embedding being generated before generating a noise image embedding before generating a noise reduced image embedding before ultimately generating the final output image from the text prompt), and wherein the diffusion prior model is trained based on training data from the first language and the second language (paras 0030, 0101: multilingual model implies multilingual (multiple languages) text embedding capabilities utilizing the text prompts to generate images via the text-to-image diffusion models as highlight in (abstract); wherein the process is carried out via processors with accompanying memory (para 0040)); and generating an image using the image generation model by performing denoising based on the image embedding (paras 0076-0085: image generator includes various steps including an image feature embedding being generated before generating a noise image embedding before generating a noise reduced image embedding before ultimately generating the final output image from the text prompt), wherein the image includes an element corresponding to the text prompt (paras 0030, 0101: multilingual model implies multilingual (multiple languages) text embedding capabilities utilizing the text prompts to generate images via the text-to-image diffusion models as highlight in (abstract); wherein the process is carried out via processors with accompanying memory (para 0040)).
Re Claim 2, Yu et al discloses the method of claim 1, further comprising: obtaining an additional text prompt in the second language (paras 0030, 0101: multilingual model implies multilingual (multiple languages) text embedding capabilities utilizing the text prompts to generate images via the text-to-image diffusion models as highlight in (abstract); wherein the process is carried out via processors with accompanying memory (para 0040)); encoding the additional text prompt using the multilingual encoder to obtain an additional multilingual text embedding (paras 0030, 0101: multilingual model implies multilingual (multiple languages) text embedding capabilities utilizing the text prompts to generate images via the text-to-image diffusion models as highlight in (abstract); wherein the process is carried out via processors with accompanying memory (para 0040)); processing the additional multilingual text embedding using the diffusion prior model to obtain an additional image embedding (paras 0076-0085: image generator includes various steps including an image feature embedding being generated before generating a noise image embedding before generating a noise reduced image embedding before ultimately generating the final output image from the text prompt); and generating an additional image using the diffusion model based on the additional image embedding, wherein the additional image includes an additional element corresponding to the additional text prompt (paras 0030, 0101: multilingual model implies multilingual (multiple languages) text embedding capabilities utilizing the text prompts to generate images via the text-to-image diffusion models as highlight in (abstract); wherein the process is carried out via processors with accompanying memory (para 0040)).
Re Claim 6, Yu et al discloses the method of claim 1, wherein: the image embedding is in a same embedding space as the multilingual text embedding (abstract, paras 0018, 0023-0024: each text prompt is embedded with an image within its own respective embedding space, wherein the text prompt and its corresponding image are embedded within the same space).
Re Claim 7, Yu et al discloses a method comprising: obtaining training data including a plurality of images, a first plurality of image captions in a first language, and a second plurality of image captions in a second language (paras 0030, 0101: multilingual model implies multilingual (multiple languages) text embedding capabilities utilizing the text prompts to generate images via the text-to-image diffusion models as highlight in (abstract); wherein the process is carried out via processors with accompanying memory (para 0040)); encoding the first plurality of image captions and the second plurality of image captions using a multilingual encoder to obtain a plurality of multilingual text embeddings (paras 0030, 0101: multilingual model implies multilingual (multiple languages) text embedding capabilities utilizing the text prompts to generate images via the text-to-image diffusion models as highlight in (abstract); wherein the process is carried out via processors with accompanying memory (para 0040)); processing the plurality of multilingual text embeddings using a diffusion prior model to obtain a plurality of predicted image embeddings corresponding to the first plurality of image captions in the first language and the second plurality of image captions in the second language (paras 0030, 0101: multilingual model implies multilingual (multiple languages) text embedding capabilities utilizing the text prompts to generate images via the text-to-image diffusion models as highlight in (abstract); wherein the process is carried out via processors with accompanying memory (para 0040)) to obtain input guidance for an image generation model different from the diffusion prior model (paras 0076-0085: image generator includes various steps including an image feature embedding being generated before generating a noise image embedding before generating a noise reduced image embedding before ultimately generating the final output image from the text prompt), and wherein the diffusion prior model is trained based on training data from the first language and the second language (paras 0030, 0101: multilingual model implies multilingual (multiple languages) text embedding capabilities utilizing the text prompts to generate images via the text-to-image diffusion models as highlight in (abstract); wherein the process is carried out via processors with accompanying memory (para 0040)); and training the diffusion prior model to generate image embeddings based on multilingual text embeddings from the first language and the second language, wherein the diffusion prior model is trained based on the plurality of predicted image embeddings and the plurality of images (paras 0030, 0101: multilingual model implies multilingual (multiple languages) text embedding capabilities utilizing the text prompts to generate images via the text-to-image diffusion models as highlight in (abstract); wherein the process is carried out via processors with accompanying memory (para 0040)).
Re Claim 11, Yu et al discloses the method of claim 7, further comprising: translating the first plurality of image captions to obtain the second plurality of image captions (paras 0030, 0101: multilingual model implies multilingual (multiple languages) text embedding capabilities utilizing the text prompts to generate images via the text-to-image diffusion models as highlight in (abstract); wherein the process is carried out via processors with accompanying memory (para 0040)).
Re Claim 12, Yu et al discloses the method of claim 7, wherein: the plurality of images includes a first subset of images corresponding to the first language and a second subset of images corresponding to the second language (paras 0030, 0101: multilingual model implies multilingual (multiple languages) text embedding capabilities utilizing the text prompts to generate images via the text-to-image diffusion models as highlight in (abstract); wherein the process is carried out via processors with accompanying memory (para 0040)), the first subset of images being different from the second subset of images (paras 0030, 0101: multilingual model implies multilingual (multiple languages) text embedding capabilities utilizing the text prompts to generate images via the text-to-image diffusion models as highlight in (abstract); wherein the process is carried out via processors with accompanying memory (para 0040)).
Re Claim 13, Yu et al discloses the method of claim 7, wherein: the multilingual encoder is pretrained prior to training the diffusion prior model (paras 0019, 0023-0024: pre-trained text models).
Claim 14 has been analyzed and rejected according to claim 1.
Re Claim 15, Yu et al discloses the system of claim 14, wherein the at least one processing device is configured to execute instructions stored in the at least one memory component to perform operations comprising: encoding, using a multilingual encoder, a text prompt in the first language to obtain the multilingual text embedding (paras 0030, 0101: multilingual model implies multilingual (multiple languages) text embedding capabilities utilizing the text prompts to generate images via the text-to-image diffusion models as highlight in (abstract); wherein the process is carried out via processors with accompanying memory (para 0040)).
Re Claim 16, Yu et al discloses the system of claim 15, wherein the multilingual encoder comprises a multimodal encoder for text and images (para 0088: datasets include multimodal evaluation dataset).
Re Claim 18, Yu et al discloses the system of claim 14, wherein: the image embedding is in a same embedding space as the multilingual text embedding (abstract, paras 0018, 0023-0024: each text prompt is embedded with an image within its own respective embedding space, wherein the text prompt and its corresponding image are embedded within the same space).
Re Claim 19, Yu et al discloses the system of claim 14, wherein the diffusion prior model comprises a transformer architecture (para 0086: transformer).
Claim Rejections - 35 USC § 103
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.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claims 3 & 20 are rejected under 35 U.S.C. 103 as being unpatentable over Yu et al, US Patent Pub. 20240185035 A1 as applied to claim 1 above, in view of Liu et al, US Patent Pub. 20230118966 A1.
Re Claim 3, Yu et al discloses the method of claim 1, but fails to disclose further comprising: generating a plurality of intermediate image embeddings corresponding to a plurality of diffusion time steps, wherein the image is generated based on the plurality of intermediate image embeddings. However, Liu et al teaches the concept of a text-to-image diffusion models where intermediate images are generated at different time intervals using a U-net for denoising corrupted images (Liu et al, para 0018). Since Yu et al discloses generating intermediate images (Yu et al, paras 0076-0085: image generator includes various steps including an image feature embedding being generated before generating a noise image embedding before generating a noise reduced image embedding before ultimately generating the final output image from the text prompt), it would have been obvious to modify Yu et al such that intermediate images at different time intervals are included as taught in Liu et al for the purpose of incorporating intermediate images in an iterative process in order to optimize and denoise as much as possible the final image.
Claim 20 has been analyzed and rejected according to claim 3.
Claim 4 is rejected under 35 U.S.C. 103 as being unpatentable over Yu et al, US Patent Pub. 20240185035 A1 as applied to claim 1 above, in view of Yu2 et al, US Patent Pub. 20240112088 A1.
Re Claim 4, Yu et al discloses the method of claim 1, but fails to disclose further comprising: obtaining a causal attention mask, wherein the image embedding is generated based on the causal attention mask. However, Yu2 et al discloses a text to image modeling system (Yu2 et al, para 0006), that is executed within a GPT style generative pretraining system (Yu2 et al, paras 0004: GPT, 0064). Since casual attention is a key mechanism used in GPT models, it would have been obvious to apply the process of Yu et al within the GPT models (along with the already incorporated casual attention masks) as taught in Yu2 et al, for the purpose of only considering the preceding tokens for prediction so that each token is predicted based on the context of the previous token.
Claim 5 is rejected under 35 U.S.C. 103 as being unpatentable over Yu et al, US Patent Pub. 20240185035 A1 as applied to claim 1 above, in view of Liu et al, US Patent Pub. 20240153152 A1.
Re Claim 5, Yu et al discloses the method of claim 1, further comprising: generating a plurality of image embeddings using the diffusion prior model (paras 0076-0085: image generator includes various steps including an image feature embedding being generated before generating a noise image embedding before generating a noise reduced image embedding before ultimately generating the final output image from the text prompt); but fails to explicitly disclose computing a similarity score between each of the plurality of image embeddings and the multilingual text embedding; and selecting the image embedding from the plurality of image embeddings based on the similarity score. However, Liu et al teaches the concept of computing a similarity score between generated image and text description and based on said similarity score, iteratively generates an image until similarity score reaches a predetermined threshold score (Liu et al, para 0017). It would have been obvious to modify the Yu et al system such that a similarity score is used to generate the final output image as taught in Liu et al for the purpose of optimizing the final output image.
Claims 8-10 & 17 are rejected under 35 U.S.C. 103 as being unpatentable over Yu et al, US Patent Pub. 20240185035 A1 as applied to claim 7 above, in view of Pang et al, US Patent Pub. 20240303764 A1.
Re Claim 8, Yu et al discloses the method of claim 7, but fails to disclose further comprising: identifying a plurality of ground-truth image embeddings corresponding to the plurality of images, respectively; and comparing the plurality of predicted image embeddings to the plurality of ground-truth image embeddings, wherein the diffusion prior model is trained based on the comparison. However, Pang et al teaches the concept of generating images from text by utilizing target ground-truth images, where the predicted images are compared to the ground-truth target image (Pang et al, para 0062) to optimize the final image (Pang et al, para 0070). It would have been obvious to modify the Yu et al system such that the images are compared to the ground truth images as taught in Pang et al for the purpose of optimizing the final image with reference to a target image.
Claim 9 has been analyzed and rejected according to claims 7-8.
Re Claim 10, the combined teachings of Yu et al and Pang et al disclose the method of claim 9, wherein: the diffusion model is pretrained prior to training the diffusion prior model (Yu et al, paras 0019, 0023-0024: pre-trained text models; Pang et al, paras 0069-0070: pre-trained diffusion model).
Claim 17 has been analyzed and rejected according to claims 7-8.
Contact
Any inquiry concerning this communication or earlier communications from the examiner should be directed to GEORGE C MONIKANG whose telephone number is (571)270-1190. The examiner can normally be reached Mon. - Fri., 9AM-5PM, ALT. Fridays off.
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, Carolyn R Edwards can be reached at 571-270-7136. 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.
/GEORGE C MONIKANG/Primary Examiner, Art Unit 2692 1/24/2026