Three layers of rights that should not be mixed

The fact that an image or a 3D model was created within Blender or DaVinci Resolve does not, by itself, resolve whether it can be used commercially. It is necessary to distinguish the software license, the terms of the model that generates the content, and the rights applicable to the result. These are distinct layers and may impose different obligations.

A model license can set conditions on using or distributing the model and its derivatives. For example, CreativeML Open RAIL-M includes use-based restrictions and sets conditions for redistribution. Exporting a result does not establish that every use is permitted.

It also matters what "commercial use" means in the specific case. Including a result in a video for a client is not the same as redistributing the model weights, selling the generated file separately, or training another model with it. Define the intended use before looking for a general answer.

Auditing a local Stable Diffusion model

In a local workflow, the starting point is to identify the exact checkpoint: the family or the name Stable Diffusion is not enough. Versions may have different licenses, and a fine-tune, a conversion, or an added component may bring additional terms. Keep the link to the specific sheet and the license that accompanies that version.

The CreativeML Open RAIL-M license of Stable Diffusion v1 describes permissions to use, reproduce, display, sublicense, and distribute the model and its derivatives, subject to conditions. It includes restrictions based on certain uses and clarifies that the person generating the result is responsible for it and its subsequent use. Commercial permission does not mean unlimited permission nor does it constitute a guarantee that each image is free of third-party rights.

The license also treats the redistribution of the model differently from the publication of a result. If you deliver or host the weights, or distribute a derivative, you must review the redistribution conditions, including the license notice and applicable restrictions. Exporting a final image is not equivalent to redistributing the model weights.

The Stability AI launch page from August 2022 documents the publication of Stable Diffusion v1 under CreativeML OpenRAIL-M. That announcement is historical and does not certify the license of other generations, revisions, or models that share the name Stable Diffusion. Always verify the specific version, not just the family name.

DaVinci Resolve and AI features

DaVinci Resolve is a tool for editing, compositing, and finishing. Its iPad listing describes AI tools powered by the DaVinci AI Neural Engine, including Magic Mask, Smart Reframe, and Voice Isolation, and says a Studio in-app upgrade is required for those features. Check the terms for the specific product and feature you use.

The DaVinci Resolve sheet for iPad identifies AI tools associated with DaVinci Neural Engine and points to Studio features such as Magic Mask, Smart Reframe, and Voice Isolation. It is a specific sheet for the iPad application: it does not allow deducing what conditions apply to each feature in desktop versions or whether a feature generates content that can be delivered as an independent asset.

For a commercial project, separate the question about which feature processes the material from the question about what rights you receive over the result. Review the current terms of Blackmagic Design for the product and platform used, in addition to the conditions of any connected external service. If the feature only analyzes or transforms existing material, do not assume it works as a generative model of Stable Diffusion.

The result does not resolve all rights

CreativeML Open RAIL-M indicates that the licensor does not claim rights over the generated result, except for the conditions established by the license itself. It also attributes responsibility for the result and its subsequent uses to the user. The license does not by itself resolve all issues regarding the use of the result.

CreativeML Open RAIL-M makes the user accountable for the generated output and its subsequent uses. It also prohibits specified uses of the model and its derivatives.

If a cloud provider generates the content, identify the service used and review its terms. CreativeML Open RAIL-M allows hosting the model for remote access by third parties, subject to distribution conditions and the usage restrictions of the license.

El resultado no resuelve todos los derechos CreativeML Open RAIL-M indica que el licenciante no reclama derechos sobre el resultado generado, salvo las condiciones que establece la...
Imagen generada con IA por BytePreview · AI generated editorial image

Checklist to document an asset before delivering it

Record the asset’s origin and the license of the model used to generate it.

  1. Note the exact model: name, version or revision, provider, and link to its license or sheet.
  2. Record the mode of use: local execution or remote access; if applicable, identify the service.
  3. Describe the intended use: material for a final work, delivery to the client, independent publication, or redistribution of weights and derivatives.
  4. Check restrictions and obligations: prohibited uses, license notices, and redistribution conditions.
  5. Keep a copy of the license and a note on the input materials.

For a team, it is advisable to associate that sheet with the final file or the project task and assign a person responsible for reviewing it before delivery. The sheet gathers the model data and its license to have them at hand during the review.

Review an AI asset before commercial delivery

A practical sequence for checking the model, service, intended use, and applicable obligations. This is not a legal determination of rights in a particular output.

  1. Model and version

    Pinpoint the exact artifact and its terms

  2. Execution and service

    Local model or hosted access

  3. Planned use

    Final work, standalone output, or model distribution

  4. License obligations

    Restrictions and distribution conditions

  5. Output review

    Consider applicable rights beyond the model license

General workflow; the Stable Diffusion v1 and DaVinci Resolve for iPad details are product-specific.

Terms vary by model revision, derivative, service, product, and platform. The sequence does not resolve third-party rights in an output.

Sources and references1

Common errors and decision criteria

A frequent error is reading "commercial" as "unrestricted". The CreativeML Open RAIL-M license allows certain uses under conditions, but retains usage restrictions and obligations for the redistribution of the model or derivatives. The second error is extending the terms of a model to all checkpoints with similar names.

It is also advisable to avoid the leap from "the application includes AI" to "the application grants all rights over any result". In DaVinci Resolve, the features mentioned for iPad describe specific Studio tools; they do not replace the review of the terms applicable to the version used or the materials that feed the process.

The practical decision can be summarized as follows: if you incorporate a result into a final work, review the model license and usage restrictions; if you redistribute the model or its derivatives, check the specific distribution conditions, including the notices and restrictions that must be communicated to those who receive it.