A material source ledger must be established during the project initiation phase.
The material sources for AIGC commercials and AI product videos are more complex than those for traditional live-action shoots. Generative imagery may include training data, prompts, reference images, style transfers, third-party model outputs, live-action footage, licensed music, fonts, and character portraits. Brands must designate a material administrator during project initiation to record the source, authorization type, scope of use, and expiration date for each piece of material. Without a ledger, subsequent acceptance and copyright tracing will fall into chaos.
The specific action is to create a material source registration form containing the material ID, name, type, source (self-built, purchased, public, generated), authorization file path, scope of authorization, restrictions, expiration date, and person responsible. This form must be confirmed at the project kickoff meeting and included as a contract attachment. The standard is that every piece of material can be traced to an authorization file or generation log, rather than relying solely on verbal explanations. The risk is that unclear material sources can render the final video unusable for commercial purposes, especially when using open-source models or third-party plugins, as authorization terms may restrict commercial use. An exception is purely original live-action footage that does not use any third-party materials, though actor portrait rights and location usage permits must still be recorded.
The brand must prepare purchase receipts, licensing contracts, free-use declarations, or terms of service for the generative tools for all intended assets. The production team must confirm whether the license for each asset covers the intended distribution channels, such as outdoor billboards, social media, television, or overseas platforms. If an asset is restricted to personal learning, it cannot be used in commercial advertising. The delivery consequence is that assets lacking authorization records will be flagged as non-compliant during acceptance and must be replaced or retroactively licensed.
During the shooting phase, metadata for both live-action and generated assets must be recorded simultaneously.
In traditional shooting, asset management revolves around camera raw files, audio files, and continuity reports. In AIGC video production, the metadata of generated assets is equally important. Every time an image or video is generated, the model version, prompt, seed value, parameter settings, generation time, and output file hash value must be recorded. This information is used to trace the generation process and to prove the originality of the assets.
The specific action is to establish a unified naming convention on set or at the generation workstation, such as Date_Scene_AssetType_VersionNumber. Live-action assets must record the shot number, shooting time, equipment parameters, and continuity notes. Generated assets must save generation logs and export them in CSV or JSON format. The standard is that every generated asset can be found in the log, and the log files are stored in the same directory structure as the asset files. The risk is that after a generative tool is updated, older model versions may become irreproducible, rendering the assets untraceable. An exception is the use of locally deployed models, but the version and checksum of the model weight files must still be recorded.
The brand must prepare shooting permits, location releases, actor likeness agreements, and prop copyright certificates. The production team must confirm the storage paths and backup strategies for generated assets to prevent log loss. If cloud generation services are used, the provider's data retention policy must also be confirmed to prevent asset deletion. The delivery consequence is that assets with incomplete metadata cannot pass copyright review, which may affect advertising placement review or platform appeals.
In the post-production phase, the licensing boundaries between generated and live-action content must be distinguished.
During post-production compositing, AIGC-generated content is often mixed with live-action assets. For example, a live-action character's background may be replaced with an AI-generated scene, or an AI-generated character may interact with a live-action actor. The brand needs to clarify which parts belong to generated content and which belong to live-action content, as the licensing foundations for the two are different. The licensing for live-action assets comes from the shooting contract, while the licensing for generated content comes from the model's terms of service and the asset sources.
The specific action is to add tags to each clip on the editing timeline, noting the asset type, source, and licensing status. Compositing nodes must save layered files to facilitate the separate export of generated or live-action layers. The standard is that every pixel in the final video can be traced back to the original asset and licensing documents. The risk is that generated content may contain third-party copyrighted elements, such as specific buildings or characters from the model's training data; even if the generated result appears original, it may trigger infringement. An exception is the use of enterprise self-trained models where all training data comes from self-owned copyrighted assets, but a training data inventory must still be recorded.
The brand must prepare licensing for style reference images of the generated content, as well as usage permissions for any trademarks or brand elements involved. The production team must confirm whether the generated content contains recognizable people, landmarks, or artworks, and obtain the corresponding licenses. If the generated content mimics a specific artist's style, an assessment must be made to determine if it constitutes infringement. The delivery consequence is that when the licensing for generated content is unclear, platforms may require the provision of generation logs or model descriptions, or else the content will be removed.
The asset licensing status must be maintained throughout version management.
AIGC video projects typically produce multiple versions, including rough cuts, fine cuts, client revisions, platform adaptations, and masters. Each version may use different asset combinations, so authorization status must be updated with each version. Brands must not only record the authorization of the final video while ignoring assets temporarily used in intermediate versions. These temporary assets may contain unauthorized test content, which poses risks if leaked or misused.
The specific action is to create an authorization snapshot for each version, recording the IDs, authorization status, and expiration dates of all assets used in that version. When updating a version, compare the old and new asset lists to remove expired or unauthorized assets. The standard is that each version can independently pass an authorization audit without relying on the final version's authorization documents. The risk is using unauthorized replacement assets during version iteration, leading to complaints against the final video. An exception is internal preview versions not used for commercial release, though their access scope must still be restricted.
Brands must prepare version management tools or naming conventions to ensure each version has a unique identifier. The production team must confirm that an asset list is automatically generated upon version export and regularly check authorization expiration dates. If asset authorization is about to expire, renewals or replacements must be arranged in advance. The delivery consequence is that if an expired asset is used during the broadcast period, the copyright holder may pursue liability, and the brand will bear the compensation costs.
The acceptance phase requires item-by-item verification of asset authorizations and generation logs.
Acceptance is the most critical phase of an AIGC video project. Brands must not only check visual effects and audio quality but also verify asset authorization status and generation logs. The acceptance checklist should include asset ledger completeness, authorization document validity, generation log readability, version consistency, platform compliance, subtitle and voiceover authorization, font copyright, music copyright, portrait rights, location permits, model service terms, data privacy statements, export file formats, and master quality.
The specific action is for acceptance personnel to check off each item, attaching the file path of the evidence for each. For example, the asset ledger corresponds to scanned authorization contracts, and the generation log corresponds to a CSV file. The standard is that each item has a clear pass or fail conclusion, with no vague descriptions. The risk is that acceptance personnel unfamiliar with AIGC technology may overlook the importance of generation logs. An exception is projects using purely live-action assets, though actor and location authorizations must still be checked.
Brands must prepare acceptance form templates and evidence file storage locations. The production team must confirm that all authorization documents and generation logs are organized and readable before acceptance. If acceptance fails, the rectification deadline and responsible party must be clearly defined. The delivery consequence is that versions failing acceptance cannot be used for commercial release; otherwise, the brand assumes legal risks.
Delivery must include complete authorization and traceability documentation.
Final delivery includes not only the finished video file but also asset authorization documents, generation logs, version notes, user manuals, and traceability reports. Brands need these documents for internal archiving, platform review, legal disputes, and subsequent derivative works. The traceability report should list the model version, prompt, seed value, and generation time for each generated asset, as well as the shooting date, location, and authorizer for each live-action asset.
The specific action is for the production team to generate an authorization and traceability list before delivery, containing all file paths and checksums. The brand must confirm that the files in the list match the final video content and back them up to secure storage. The standard is that third-party auditors can trace the source and authorization of each asset using only the list and files. The risk is that incompatible file formats or damaged storage media may lead to the loss of traceability information. An exception is projects used solely for internal display and not published externally, though retaining basic records is still recommended.
The brand must prepare storage space and file naming conventions. The production team must confirm that deliverables are stored in directories by type and provide an index file. The consequence of delivery is that without traceability documents, platform review may fail, or evidence cannot be provided for subsequent rights protection.
Applicable Boundaries and Inapplicable Scenarios
This method for recording material sources and authorizations applies to most AIGC commercial and AI product video projects, but not all projects require the same depth. If the project is purely live-action and uses no generative technology, the material ledger can be simplified, but actor and location authorizations must still be recorded. If the project consists entirely of generated content using public commercial services, the focus must be on recording the model's terms of service and generation logs.
Inapplicable scenarios include quick test videos for internal brainstorming that do not involve commercial release, which do not require a complete ledger. Another inapplicable scenario is the use of fully open-source models with no commercial restrictions, but even then, it is recommended to retain generation logs for verification. The risk is that the brand may mistakenly believe all AIGC content requires no authorization, leading to infringement disputes. An exception is the use of an enterprise's self-built model trained entirely on proprietary data, but proof of the training data source must be provided.
When evaluating a project, the brand must judge the complexity of the material sources and the level of authorization risk. If the project involves a large amount of third-party materials or generative models, a complete recording process must be established. If the project is simple, it can be appropriately simplified, but key authorization documents cannot be missing. The production team must confirm the depth of recording with the brand at project kickoff to avoid difficulties with retroactive recording later.
Next Steps
It is recommended that before starting an AIGC commercial or AI product video project, the brand hold a special meeting on material authorization with the production team to clarify the ledger format, recording responsibilities, acceptance criteria, and delivery checklist. Clauses regarding material authorization and traceability can be added to the contract, requiring the production team to provide complete documentation upon delivery. If the project involves complex generated content, it is recommended to consult legal professionals to evaluate the model's terms of service and training data risks. Maintain a habit of record-keeping; even for small-scale projects, it provides a foundation for subsequent version iterations and rights protection.
If you are preparing an AIGC commercial project, you can first organize the brief, reference visuals, product or corporate materials, delivery platforms, and copyright scope, and then view theAIGC video service pageto translate communication from abstract preferences into executable production boundaries.
