# US 2026/0148134 A1 · Method, System, and Computer-Readable Medium for Training and Deploying a Model to Analyze and Generate Video Content Based on Cinematic Elements Using Metadata from a Captioner AI

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2026/0148134 A1
- Document ID: US20260148134A1
- Status: Published application
- Kind: A1
- Application number: 19/078,242
- Filed: 2025-03-12
- Published: 2026-05-28
- Earliest claimed: 2024-11-27
- Assignees: InterPositive, LLC
- Applicants: InterPositive, LLC
- Inventors: Benjamin Geza Affleck-Boldt
- Web page: https://patentgrep.com/patent/US20260148134A1

## CPC

- G06V 20/49: {Segmenting video sequences, i.e. computational techniques such as parsing or cutting the sequence, low-level clustering or determining units such as shots or scenes}
- G06N 20/00: Machine learning
- G06V 10/32: Normalisation of the pattern dimensions
- G06V 10/774: Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
- G06V 10/82: using neural networks
- G06V 20/41: {Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items (segmenting video sequences G06V20/49)}
- G06V 20/46: {Extracting features or characteristics from the video content, e.g. video fingerprints, representative shots or key frames}
- G06V 20/70: Labelling scene content, e.g. deriving syntactic or semantic representations
- G11B 27/031: Electronic editing of digitised analogue information signals, e.g. audio or video signals
- H04N 21/4666: {using neural networks, e.g. processing the feedback provided by the user}
- H04N 21/816: {involving special video data, e.g 3D video}
- H04N 21/84: Generation or processing of descriptive data, e.g. content descriptors {(systems specially adapted for using meta-information in broadcast systems H04H60/73)}
- G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G11B: INFORMATION STORAGE BASED ON RELATIVE MOVEMENT BETWEEN RECORD CARRIER AND TRANSDUCER (recording measured values in a way that does not require playback through a transducer G01D9/00; recording or playback apparatus using mechanically marked tape, e.g. punched paper tape, or using unit records, e.g. punched or magnetically marked cards G06K; transferring data from one type of record carrier to another G06K1/18; circuits for coupling output of reproducer to radio receiver H04B1/20; gramophone pick-ups or like acoustic electromechanical transducers or circuits therefor H04R)
- H04N: PICTORIAL COMMUNICATION, e.g. TELEVISION

## Abstract

A method trains a new model to analyze and generate video content using a captioner AI by accessing a trained model, compiling a dataset, generating metadata, designing a new model architecture, training the new model with metadata, validating accuracy, refining the model, deploying the model, and updating the training dataset. A computing trains a new model to analyze and generate video using a captioner AI by accessing a trained model, compiling a dataset, generating metadata, designing a new model architecture, training the new model with metadata, validating accuracy, refining the model, deploying the model, and updating the training dataset. A computer-readable medium trains a new model to analyze and generate video using a captioner AI by accessing a trained model, compiling a dataset, generating metadata, designing a new model architecture, training the new model with metadata, validating accuracy, refining the model, deploying the model, and updating the training dataset.

## Claims

### Claim 1

A computer-implemented method for training a model to analyze and generate video content based on cinematic elements using a captioner AI, the method comprising:
a. accessing a trained captioner AI model capable of identifying and generating metadata for cinematic elements within video sequences, where the cinematic elements include focal length, camera movement, depth of field, occlusion, and parallax;
b. compiling a dataset of video sequences that have not been previously analyzed by the captioner AI model, where each video sequence includes a variety of cinematic elements;
c. processing the compiled dataset through the trained captioner AI model to generate metadata for each video sequence, where the metadata describes the cinematic elements present in the sequence;
d. defining and instantiating a model architecture including a visual feature extractor, a temporal sequence model, and a fusion layer that embeds metadata into a feature vector and concatenates the metadata features with visual features to generate outputs that control generation of video sequences based on input cinematic parameters;
e. training the model using the metadata generated by the captioner AI model as training data, where the training process conditions the model to recognize and replicate the identified cinematic elements in video content;
f. validating an accuracy of the model in recognizing and generating cinematic elements within video sequences using a separate validation dataset processed through the captioner AI model;
g. iteratively refining the model based on performance feedback obtained during validation to enhance its ability to accurately analyze and generate video content based on cinematic elements;
h. deploying the trained model to analyze incoming video content and generate video sequences that adhere to specified cinematic parameters; and
i. continuously updating the training dataset for the model with additional metadata generated by the captioner AI model from newly compiled video sequences to ensure the model remains current with cinematic trends and techniques.

### Claim 2 (depends on 1)

The computer-implemented method of claim 1, wherein the trained captioner AI model has been fine-tuned using a dataset that includes video sequences across multiple cinematic styles and genres and predefined shot types and lighting setups, to enable generation of the metadata.

### Claim 3 (depends on 1)

The computer-implemented method of claim 1, wherein the model architecture includes one or more fully-connected layers that process metadata into an embedding and a concatenation layer that combines metadata features to form a fused representation used for prediction.

### Claim 4 (depends on 1)

The computer-implemented method of claim 1, wherein validating the accuracy includes computing a visual quality loss and one or more cinematic parameter losses including at least a mean squared error for focal length predictions and a classification accuracy for shot type predictions on the validation dataset processed through the captioner AI model.

### Claim 5 (depends on 1)

The computer-implemented method of claim 1, wherein validating the accuracy of the model includes comparing the model's analysis of video content with the metadata generated by the captioner AI model to identify discrepancies and areas for improvement.

### Claim 6 (depends on 1)

The computer-implemented method of claim 1, further comprising utilizing the model in real-time video production environments to provide immediate analysis and feedback on a cinematic quality of captured video content.

### Claim 7 (depends on 1)

The computer-implemented method of claim 1, wherein the continuous updating of the training dataset includes processing newly released films and video content through the captioner AI model to capture evolving cinematic styles and techniques.

### Claim 8 (depends on 1)

The computer-implemented method of claim 1, further comprising integrating the model into video editing software to automate the generation of video sequences that comply with user-specified cinematic parameters.

### Claim 9 (depends on 1)

The computer-implemented method of claim 1, wherein deploying the trained model includes making the model available as a cloud-based service, where users can upload video content for analysis and request the generation of new video sequences based on desired cinematic elements.

### Claim 10 (depends on 1)

The computer-implemented method of claim 1, further comprising implementing a feedback mechanism where users can rate the accuracy and quality of the model's analysis and generated video content, providing valuable data for ongoing refinement of the model.

### Claim 11

A computing system for training a model to analyze and generate video content based on cinematic elements using a captioner AI, the computing system comprising:
one or more processors; and
one or more memories, having stored thereon instructions that, when executed, cause the computing system to:
access a trained captioner AI model capable of identifying and generating metadata for cinematic elements within video sequences,
where the cinematic elements include focal length, camera movement, depth of field, occlusion, and parallax; compile a dataset of video sequences that have not been previously analyzed by the captioner AI model,
where each video sequence includes a variety of cinematic elements;
process the compiled dataset through the trained captioner AI model to generate metadata for each video sequence,
where the metadata describes the cinematic elements present in the sequence;
define and instantiate a model architecture including a visual feature extractor, a temporal sequence model, and a fusion layer that embeds metadata into a feature vector and concatenates the metadata features with visual features to generate outputs that control generation of video sequences based on input cinematic parameters;
train the model using the metadata generated by the captioner AI model as training data,
where the training process conditions the model to recognize and replicate the identified cinematic elements in video content;
validate an accuracy of the model in recognizing and generating cinematic elements within video sequences using a separate validation dataset processed through the captioner AI model;
iteratively refine the model based on performance feedback obtained during validation to enhance its ability to accurately analyze and generate video content based on cinematic elements;
deploy the trained model to analyze incoming video content and generate video sequences that adhere to specified cinematic parameters; and
continuously update the training dataset for the model with additional metadata generated by the captioner AI model from newly compiled video sequences to ensure the model remains current with cinematic trends and techniques.

### Claim 12 (depends on 11)

The computing system of claim 11, the memories having stored thereon further instructions that, when executed, cause the computing system to ensure the trained captioner AI model has been fine-tuned using a dataset that includes video sequences across multiple cinematic styles and genres and predefined shot types and lighting setups, to enable generation of the metadata.

### Claim 13 (depends on 11)

The computing system of claim 11, the memories having stored thereon further instructions that, when executed, cause the computing system to include in the model architecture one or more fully-connected layers that process metadata into an embedding and a concatenation layer that combines metadata features to form a fused representation used for prediction.

### Claim 14 (depends on 11)

The computing system of claim 11, the memories having stored thereon further instructions that, when executed, cause the computing system to validate the accuracy of the model by computing a visual quality loss and one or more cinematic parameter losses including at least a mean squared error for focal length predictions and a classification accuracy for shot type predictions on the validation dataset processed through the captioner AI model.

### Claim 15 (depends on 11)

The computing system of claim 11, the memories having stored thereon further instructions that, when executed, cause the computing system to validate the accuracy of the model by comparing the model's analysis of video content with the metadata generated by the captioner AI model to identify discrepancies and areas for improvement.

### Claim 16 (depends on 11)

The computing system of claim 11, the memories having stored thereon further instructions that, when executed, cause the computing system to utilize the model in real-time video production environments to provide immediate analysis and feedback on a cinematic quality of captured video content.

### Claim 17 (depends on 11)

The computing system of claim 11, the memories having stored thereon further instructions that, when executed, cause the computing system to continuously update the training dataset by processing newly released films and video content through the captioner AI model to capture evolving cinematic styles and techniques.

### Claim 18 (depends on 11)

The computing system of claim 11, the memories having stored thereon further instructions that, when executed, cause the computing system to integrate the model into video editing software to automate the generation of video sequences that comply with user-specified cinematic parameters.

### Claim 19 (depends on 11)

The computing system of claim 11, the memories having stored thereon further instructions that, when executed, cause the computing system to deploy the trained model as a cloud-based service, where users can upload video content for analysis and request the generation of new video sequences based on desired cinematic elements.

### Claim 20 (depends on 11)

The computing system of claim 11, the memories having stored thereon further instructions that, when executed, cause the computing system to implement a feedback mechanism where users can rate the accuracy and quality of the model's analysis and generated video content, providing valuable data for ongoing refinement of the model.

### Claim 21

A computer-readable medium having stored thereon a set of non-transitory instructions that, when executed, cause a computer to:
access a trained captioner AI model capable of identifying and generating metadata for cinematic elements within video sequences,
where the cinematic elements include focal length, camera movement, depth of field, occlusion, and parallax;
compile a dataset of video sequences that have not been previously analyzed by the captioner AI model,
where each video sequence includes a variety of cinematic elements;
process the compiled dataset through the trained captioner AI model to generate metadata for each video sequence,
where the metadata describes the cinematic elements present in the sequence;
define and instantiate a model architecture including a visual feature extractor, a temporal sequence model, and a fusion layer that embeds metadata into a feature vector and concatenates the metadata features with visual features to generate outputs that control generation of video sequences based on input cinematic parameters;
train the model using the metadata generated by the captioner AI model as training data,
where the training process conditions the model to recognize and replicate the identified cinematic elements in video content;
validate an accuracy of the model in recognizing and generating cinematic elements within video sequences using a separate validation dataset processed through the captioner AI model;
iteratively refine the model based on performance feedback obtained during validation to enhance its ability to accurately analyze and generate video content based on cinematic elements;
deploy the trained model to analyze incoming video content and generate video sequences that adhere to specified cinematic parameters; and
continuously update the training dataset for the model with additional metadata generated by the captioner AI model from newly compiled video sequences to ensure the model remains current with cinematic trends and techniques.

### Claim 22 (depends on 21)

The computer-readable medium of claim 21 having stored thereon further instructions that, when executed, cause a computer to fine-tune the trained captioner AI model using a dataset that includes video sequences across multiple cinematic styles and genres and predefined shot types and lighting setups, to enable generation of the metadata.

### Claim 23 (depends on 21)

The computer-readable medium of claim 21 having stored thereon further instructions that, when executed, cause a computer to include in the model architecture one or more fully-connected layers that process metadata into an embedding and a concatenation layer that combines metadata features to form a fused representation used for prediction.

### Claim 24 (depends on 21)

The computer-readable medium of claim 21 having stored thereon further instructions that, when executed, cause a computer to validate the accuracy of the model by computing a visual quality loss and one or more cinematic parameter losses including at least a mean squared error for focal length predictions and a classification accuracy for shot type predictions on the validation dataset processed through the captioner AI model.

### Claim 25 (depends on 21)

The computer-readable medium of claim 21 having stored thereon further instructions that, when executed, cause a computer to validate the accuracy of the model by comparing the model's analysis of video content with the metadata generated by the captioner AI model to identify discrepancies and areas for improvement.

### Claim 26 (depends on 21)

The computer-readable medium of claim 21 having stored thereon further instructions that, when executed, cause a computer to utilize the model in real-time video production environments to provide immediate analysis and feedback on a cinematic quality of captured video content.

### Claim 27 (depends on 21)

The computer-readable medium of claim 21 having stored thereon further instructions that, when executed, cause a computer to continuously update the training dataset by processing newly released films and video content through the captioner AI model to capture evolving cinematic styles and techniques.

### Claim 28 (depends on 21)

The computer-readable medium of claim 21 having stored thereon further instructions that, when executed, cause a computer to integrate the model into video editing software to automate the generation of video sequences that comply with user-specified cinematic parameters.

### Claim 29 (depends on 21)

The computer-readable medium of claim 21 having stored thereon further instructions that, when executed, cause a computer to deploy the trained model as a cloud-based service, where users can upload video content for analysis and request the generation of new video sequences based on desired cinematic elements.

### Claim 30 (depends on 21)

The computer-readable medium of claim 21 having stored thereon further instructions that, when executed, cause a computer to implement a feedback mechanism where users can rate the accuracy and quality of the model's analysis and generated video content, providing valuable data for ongoing refinement of the model.
