Have you tried VQGAN+CLIP? https://github.com/nerdyrodent/VQGAN-CLIP

You can also use it through your browser via Google Colab: https://colab.research.google.com/drive/1Bzyn7cveZvc3_KHteje...