# Provide CLIP embeddings directly to diffuser

**URL:** </static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fprovide-clip-embeddings-directly-to-diffuser%2F49614%3E%5C
**Category:** Beginners\
**Created:** [August 5, 2023, 8:41pm UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fprovide-clip-embeddings-directly-to-diffuser%2F49614 "2023-08-05T20:41:43Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![guyross](https://avatars.discourse-cdn.com/v4/letter/g/a698b9/32.png) [@guyross](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2Fguyross)%5C
**Post date:** [August 5, 2023, 8:41pm UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fprovide-clip-embeddings-directly-to-diffuser%2F49614%2F1 "2023-08-05T20:41:43Z")

</div>

Is there any way to provide the embeddings directly to the diffusion pipeline? I specifically want to provide CLIP embeddings coming from a different process, directly to the diffusion pipeline, in place of the text prompt, but I could not find any simple way to do it in the documentation.

Edit:  
After investigating the code I noticed the “prompt\_embeds” arg exists on the pipeline, but im still trying to pass in image embeddings from CLIP.

E.g  
pipeline(prompt=“a dog sitting on a bench”, image=img, mask\_image=mask\_image).images[0]

Instead of the “prompt”, I can provide clip embeddings.