Commercial use: read and record each model, LoRA and node licence before client work

Purpose and scope

Read and record the licence of every model, LoRA, custom node and workflow before client or commercial work. This guide summarises only what each text states for the model families this site covers. It is not legal advice.

Source-supported context

What the texts say on 2026-10-06:

• SDXL base 1.0 is under the CreativeML Open RAIL++-M License dated July 26, 2023. Section III covers hosting for third-party remote access and redistribution, which must pass on the use-based restrictions, a licence copy and notices; it makes the Attachment A restrictions binding and states that the licensor claims no rights in the Output you generate and that you are accountable for it. Source 1
• FLUX.1 [dev] is under the FLUX.1 [dev] Non-Commercial License v1.1.1. Section 2(b) allows access, use, distribution and derivatives only for Non-Commercial Purposes and says other uses, such as commercial activity, need a licence from Black Forest Labs. Section 1(c) defines Non-Commercial Purpose and excludes revenue-generating activity and direct interaction with end users. Section 2(d) says the company claims no ownership of Outputs and that you may use Output for any purpose, including commercial, except as expressly prohibited, and not to train a model competitive with FLUX.1 [dev] or FLUX.1 Kontext [dev]. Section 2(e) conditions use on content filtering or review of Output and, where law requires, an AI disclosure. Section 3 sets Distribution conditions; Section 4 lists restrictions, and 4(a) bars use of the model, derivatives "or any data produced by the" model for commercial or production purposes. The text does not reconcile 4(a) with 2(d); ask Black Forest Labs or counsel. Source 2
• FLUX.1 [schnell]: the model card states apache-2.0 and says the model can be used for personal, scientific and commercial purposes. Source 3
• Qwen-Image (the original model): the card has a "License Agreement" heading stating Qwen-Image is licensed under Apache 2.0. Source 4 Qwen-Image-2.1 is a different release with a different licence: its repository says it is under the Qwen Research License Agreement (release date September 20, 2026), whose Section 2(a) grants rights for non-commercial purposes only and Section 2(b) requires a separate commercial licence from the licensor. Source 8 Other Qwen-Image variants have their own cards; read each one.
• Wan 2.2 (T2V-A14B card): the card states Apache 2.0, says the authors claim no rights over generated content, and says use must not involve sharing content that violates law or causes harm. Source 5
• Apache 2.0 (behind schnell, Qwen-Image and Wan 2.2): redistribution needs a licence copy, change notices and retained notices (Section 4); no trademark grant (Section 6); provided "AS IS" (Section 7). The text does not mention generated outputs. Source 7
• ComfyUI itself is under GPL-3.0. Section 2 affirms unlimited permission to run the unmodified Program and says output is covered by the licence only if it constitutes a covered work; Section 4 allows charging any price for conveyed copies; conveying modified versions requires source under the same licence (Sections 5-6). Source 6

Before you start

1. List every model, encoder, VAE, LoRA, ControlNet and custom node in the workflow.

2. Note each one's repository and revision, since the licence belongs to that revision.

3. Know what the client contract requires on output ownership.

Steps and checks

The sequence below is an editorial procedure based on the cited context, not a diagnosis already confirmed for your environment.

Step 1. Open each licence file or card at its source and save a dated copy with the URL and revision.

Step 2. Record the licence name and version, and whether commercial use is stated as permitted, restricted or not addressed, with the clause number.

Step 3. Record separately what is said about outputs, redistribution, attribution and acceptable use; these can differ from the commercial-use answer.

Step 4. Check whether the artifact is a finetune, merge, LoRA or converted file; it can carry its own licence.

Step 5. Check each custom node against its own repository licence, not ComfyUI's.

Step 6. Resolve every "restricted" or "unclear" item with the rights holder or qualified counsel before delivery.

Step 7. Re-check at each model update and new client project, and store the check date.

Completion check

Every artifact has a licence name, clause reference, saved copy and check date.

Limitations and cautions

This is not legal advice; jurisdiction and contracts matter. Licences can change, and this page reflects 2026-10-06 readings read from the original licence files and model cards, so confirm against the current text. The Hugging Face copy of the FLUX.1 [dev] licence is gated and was not read; the Black Forest Labs GitHub copy was used. A base-model licence does not extend to a LoRA or finetune built on it. Other Stable Diffusion releases and training data are not covered.

Original sources

• SDXL base 1.0 licence — checked 2026-10-06.
• FLUX.1 [dev] licence (Black Forest Labs GitHub copy) — checked 2026-10-06.
• FLUX.1 [schnell] model card — checked 2026-10-06.
• Qwen-Image model card — checked 2026-10-06.
• Wan2.2 model card — checked 2026-10-06.
• ComfyUI LICENSE — checked 2026-10-06.
• Apache License 2.0 text — checked 2026-10-06.
• Qwen-Image-2.1 LICENSE — checked 2026-10-06.

Prepared 2026-10-06 from the sources above. No GPU run, quality measurement or runtime guarantee is supplied.

Related troubleshooting and guides

• Record model downloads by repository, revision, path and integrity evidence
• Add a LoRA to ComfyUI: verify the file and both model/CLIP branches
• Install a custom node in three stages: code, dependencies and registration