Tutorial: Image Editing with Midjourney and Nano-Banana (gbti.network)

This tutorial demonstrates a practical, end-to-end workflow for fixing garbled text in MidJourney images by combining MidJourney’s editor with the inpainting tool Nano-Banana. MidJourney excels at creative composition but routinely fails to render readable text; the author generated variations, then used Nano-Banana to correct the text area. Because Nano-Banana introduced problems—loss of font styling, incorrect sizing, a watermark, and some quality degradation—the image was then brought back into MidJourney’s editor for targeted fixes. Using the eraser plus prompt-guided generative-fill variations, they replaced the bad text with “Tutorial,” removed the Nano-Banana watermark, adjusted composition, and upscaled the final asset in MidJourney’s Gallery to recover image quality. The whole pipeline took about an hour. For the AI/ML community this is a useful case study in tool interoperability and practical trade-offs when chaining generative models: inpainting services can correct localized errors but may change style, introduce artifacts/watermarks, or reduce fidelity, requiring iterative edits and upscaling. Technical takeaways: use prompt-guided erasure and variation selection in MidJourney for fine corrections; quote words you must preserve when retexturizing to reduce hallucinated changes; expect style-loss and plan for final upscaling. Ultimately, improvements in model editing could eliminate the need for such multi-tool workflows, but today this approach offers a pragmatic path to salvage text-heavy compositions.