No audio on Open WebUI + Kokoro-FastAPI

[RBEmerson970]

I have a working Kokoro-FastAPI installation working under Open WebUI under Docker under Win 11. Using localhost:3003 works - that is, I see a functioning page which responds to supplying a PDF, and opening it.

OpenReader moves through the pages of a PDF doc, starting with a greyed lines at the top, ending with a greyed line at the bottom, but no audio.

I've tried this with Kokoro-FastAPI running and not running - no difference. The installation appears to have run OK:

`PS C:\Users\pavil> docker run --name openreader-webui -p 3003:3003 richardr1126/openreader-webui:latest
Unable to find image 'richardr1126/openreader-webui:latest' locally
latest: Pulling from richardr1126/openreader-webui
95d6a69336c7: Download complete
3278698e8085: Download complete
b4d603fdea2d: Download complete
e45b504d77a5: Download complete
d8656c5f9e63: Download complete
a101b4c87ad9: Download complete
01d80474c708: Download complete
e64437407f53: Download complete
729348fc3970: Download complete
Digest: sha256:5bfd5390ce0f5941d9757b9c309523b420df1111714af9d0126dcd0e46f35c4f
Status: Downloaded newer image for richardr1126/openreader-webui:latest

> openreader-webui@0.1.0 start
> next start -p 3003

   ▲ Next.js 15.1.5
   - Local:        http://localhost:3003
   - Network:      http://172.17.0.3:3003

 ✓ Starting...
 ✓ Ready in 275ms

The list of voices doesn't fit the form l{f|m}_name and it certainly doesn't have all the names available through Kokoro-FastAPI.
I also tried using --gpus all, but that seems to have failed. I deleted the container, and installed without the --gpus - no change.

[richardr1126]

The project should work with Kokoro-FastAPI.

Are you setting the correct API base url in the settings? If both Kokoro-FastAPI and my app are in docker containers, then you might want to try setting the Kokoro base URL to http://host.docker.internal:8880/v1, if you already tried http://localhost:8880/v1.

The error you are getting is because your API is not responding, so it keeps skipping to the next sentence.

My public live demo is currently using Kokoro-FastAPI v0.2.0, just updated it yesterday.

I have now tested the Docker container and can confirm it works using a local Kokoro-FastAPI on http://localhost:8880/v1

[RBEmerson970]

Setting OpenAI API Base URL to
http://host.docker.internal:8880/v1
fixed the problem.
Here's the Docker situation:

[richardr1126]

👍 I thought that could be the problem as I have had that problem before

I tested it with host.docker.internal for myself and that didn't work on my end so I edited my answer lol. Should've stuck with my gut.

[RBEmerson970]

I think this thread can be closed. :)