A/B AI Image Edit Comparison Workflow for n8n + Slack

Run one AI image edit through two models—and compare the results in Slack

This A/B AI Image Edit Comparison workflow for n8n sends the same prompt and same reference image to two AI image-editing models in parallel, then posts both outputs side-by-side in Slack. Perfect for evaluating quality before you commit to production.

What this workflow does

• Collects inputs in an n8n form: an edit instruction (prompt) plus a public reference image URL. There’s no chat UI—runs can also be triggered later via webhook or schedule.
• Builds an A/B render flow via CNAPS MCP: an AI Agent calls the CNAPS MCP server to register the image and fan out the same prompt to two model IDs taken from the Settings node.
• Waits for both renders to finish: a Wait node holds while the models render. Polling occurs outside the agent to avoid rate-limit risk and keep runs efficient.
• Normalizes results: a second AI Agent reads flow status back from CNAPS, and a Code node converts it into a single payload containing status, the model IDs that ran, and one output URL per model.
• Validates completion before posting: an IF node checks whether both models returned an image. If one model is empty, the workflow reports the failure instead of posting a false success.
• Posts to Slack: sends either both output URLs for side-by-side comparison or a failure notice naming which model came back empty.

Use cases

• Prompt quality testing for image-editing changes before rolling out to customers.
• Model comparison when swapping image-editing model IDs is “just a text edit” in settings.
• Automated evaluation for SaaS operators who need consistent A/B checks and Slack notifications.

Technical details

• Integrations: CNAPS MCP, Slack
• n8n nodes: if, set, code, wait, merge, slack