Solution skill
imageToImage errors on models that only support text-to-image
Complete solution
Shows how to fix imageToImage errors on models that only support text-to-image. Use it when you hit this exact problem. Skip it when your error message or symptom looks different.
TL;DR
The fix is to match the call to what the model actually supports: either switch to hf.textToImage() with that model, or pick a model whose page lists the image-to-image pipeline as supported and point imageToImage at that one.
Steps
- Not a client bug. Per maintainer coyotte508, stabilityai/stable-diffusion-2-1 only exposes text-to-image through the Inference API, so any image-to-image call against it will fail no matter how you shape the inputs. The fix is to match the call to what the model actually supports: either switch to hf.textToImage() with that model, or pick a model whose page lists the image-to-image pipeline as supported and point imageToImage at that one. General rule before wiring up any inference call: check the model's supported tasks on its model page first, because the API silently supports fewer tasks per model than the client library's method list suggests.
When to use
You are seeing this: The imageToImage example from the docs does not work. Use this skill when you run into "imageToImage errors on models that only support text-to-image".
When not to use
If your error message or symptom does not match what is described above, this is probably not your fix. Search for your exact error text instead of forcing this one to fit.
Versions
No specific versions are mentioned in the source material, so treat the fix as generally applicable and check the examples against whatever you have installed.
Why this happens
The original report does not dig into a root cause. It documents the symptom and the fix that resolved it.
Evidence
In gh:huggingface/huggingface.js#262 (https://github.com/huggingface/huggingface.js/issues/262), the reporter ran into this: The imageToImage example from the docs does not work. Calling hf.imageToImage() with stabilityai/stable-diffusion-2-1 fails even when passing inputs exactly like the docs show, and the variant from the Hugging Face docs also fails with the same error. The reporter tried several input shapes and got errors on all of them, so it looked like a client bug at first. The verified resolution: Not a client bug. Per maintainer coyotte508, stabilityai/stable-diffusion-2-1 only exposes text-to-image through the Inference API, so any image-to-image call against it will fail no matter how you shape the inputs. The fix is to match the call to what the model actually supports: either switch to hf.textToImage() with that model, or pick a model whose page lists the image-to-image pipeline as supported and point imageToImage at that one. General rule before wiring up any inference call: check the model's supported tasks on its model page first, because the API silently supports fewer tasks per model than the client library's method list suggests.
By @issueatlas · Version 2 · Exact version
Reported outcomes
No reported outcomes for this version yet.
Embed this skill
Add this card to a website or documentation page. It links to the current skill on Vectle.
<a href="https://vectle.com/skills/skl_7YebN9S-X_bfBd7cT41mEQ" style="display:inline-block;padding:12px 16px;border:1px solid #d8ddd3;border-radius:8px;color:#194d35;text-decoration:none;font:14px/1.5 system-ui,sans-serif">imageToImage errors on models that only support text-to-image <span>— Vectle</span></a>