Produce marketing creative variants with ComfyUI: keep a manifest, approval gates and platform AI rules before release
Advertising teams generating many variants through the ComfyUI API need a per-variant record linking parameters, prompt_id, output file and approver. This guide sets out that structure and points to the FTC, Google and Meta pages (read 2026-10-06) relevant to AI-generated or edited ad imagery; it is not legal advice.
Purpose and scope
Marketing teams need many traceable, reviewable variants of one approved concept. This guide gives a record and approval structure around ComfyUI's API. It is not legal advice; rules depend on jurisdiction, channel and contracts.
Source-supported context
API-format JSON omits UI metadata; the page's example shows nodes keyed by numeric ID with class_type and inputs, where widget values such as seed appear as literals and links to other nodes appear as ["4", 0]-style arrays (changing the literals is this guide's editorial choice) (Source 1). POST to /prompt returns a prompt_id, /history/{prompt_id} retrieves that execution, and /view serves images (Source 2). Embedded metadata is not a digital signature and does not prove who created or modified a file, a file re-encoded by another application may no longer contain it, and it should be treated as untrusted input (Source 3).
The FTC's reviews and testimonials rule (announced 2024-08-14) covers fake reviews by people who do not exist, such as AI-generated ones (Source 5); its endorsement guidance asks for disclosure of material connections (Source 4). Google Ads does not allow manipulating media to deceive, defraud or mislead others, and lists deceptively doctoring media about politics, social issues or matters of public concern as an example (Source 7); its AI labelling update, posted 2026-07-09, says AI regulations in the European Union, India and New York require disclosures or labels on ads with certain AI-generated or edited assets, permits labels inside image and video creatives, and says use of its AI label setting does not guarantee compliance with specific regulations (Source 6). Meta says ads fully created or materially edited with its own generative AI creative features carry "AI info", gives resizing and colour correction as minor changes that do not, and lists ads created or modified with third-party creation or AI tools among those that will not have it; separately, for ads about social issues, elections or politics, advertisers must already disclose digitally created or altered image, video or audio, including through third-party AI tools (Source 8).
Before you start
1. An approved workflow that already runs for one variant.
2. Brand guidelines as reviewable checks, with named approvers.
3. Record storage and a channel list.
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. Fix the template. Export the approved workflow in API format, version it, and decide which inputs vary (prompt, seed, input image) and which are frozen.
Step 2. Keep a manifest with one row per variant (ID, channel, prompt fields, seed, input asset) and write each row into a copy of the template's literal inputs, never its links.
Step 3. Submit each copy to /prompt and store the returned prompt_id in the manifest row at once; record submission errors as their own state.
Step 4. Look up each prompt_id in history, download files through /view, and store hash, filename and manifest row together; do not rely on embedded metadata alone.
Step 5. Gate before release: brand and legal or claims review, each with approver, decision and date.
Step 6. For each channel, note whether AI labelling, manipulated-media or endorsement rules apply, with the policy page and date read.
Step 7. Archive per variant: submitted JSON, model identities, manifest row, approvals, delivered file.
Completion check
Each delivered variant maps to a manifest row, prompt_id, JSON, file hash, approver and dated policy reading.
Limitations and cautions
Policy pages change; only the Google AI labelling page shows a date (posted 2026-07-09), and the other Google, Meta, FTC and ComfyUI pages show no edit date. Re-read them per campaign. A missing Meta AI info label on an ad made with third-party tools is not evidence that no disclosure duty applies. Meta describes its labelling as being gradually introduced. Seed reproducibility was not tested. The FTC pages read here do not address a general duty to label AI imagery, and the fake-reviews rule concerns reviews and testimonials, not ad imagery generally; ask counsel.
Original sources
- ComfyUI API format — checked 2026-10-06.
- ComfyUI server routes — checked 2026-10-06.
- ComfyUI metadata — checked 2026-10-06.
- FTC endorsements guidance — checked 2026-10-06.
- FTC fake reviews rule — checked 2026-10-06.
- Google Ads AI labelling, July 2026 — checked 2026-10-06.
- Google Ads manipulated media — checked 2026-10-06.
- Meta AI info labels — 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
Was this page helpful?
Anonymous. Only a yes/no counter is stored; no account, IP address or device details.
Sources & references
Editorial research from the cited public sources, dated 2026-10-06. This site did not run or test the workflow described.
01ComfyUI workflow API formatSource checked: 2026-10-0602ComfyUI server routes and messagesSource checked: 2026-10-0603ComfyUI workflow metadataSource checked: 2026-10-0604FTC: endorsements, influencers and reviewsSource checked: 2026-10-0605FTC press release: final rule on fake reviews and testimonialsSource checked: 2026-10-0606Google Ads: updates to AI labelling requirements (July 2026)Source checked: 2026-10-0607Google Ads: manipulated mediaSource checked: 2026-10-0608Meta: AI-generated images in ads, AI info labelsSource checked: 2026-10-06Report an issue · 9e704d2b-9b4b-5575-b667-f772a0959390
Related reading
Pages linked to this one by the editors.
This guide concerns an instance you own or are authorized to use. ComfyUI's native HTTP API is distinct from this site's knowledge MCP and does not authorize access to third-party services.
After execution, identify which nodes produced results and where those results reside. Node IDs, prompt IDs and filenames are different identifiers.
A model name can cover bundled checkpoints, split weights, quantized variants, text encoders and VAEs. Preserve enough identity information to know which artifact a workflow actually used.
Identify the format and source of JSON, images or videos before installing nodes and weights. Static inspection helps planning but cannot replace execution in the target environment.