Embedded model fails to load from paths with non-ASCII characters

[bpasero]

IN ORDER TO ASSIST YOU, PLEASE PROVIDE THE FOLLOWING:

• Speech SDK log taken from a run that exhibits the reported issue.

log.txt

• A stripped down, simplified version of your source code that exhibits the issue. Or, preferably, try to reproduce the problem with one of the public samples in this repository (or a minimally modified version of it), and share the code.

// the path below needs to exist on disk and needs to contain spaces and non-ASCII characters
auto speechConfig = EmbeddedSpeechConfig::FromPath("/Users/bpasero/Desktop/assets sind cööl/stt");

Describe the bug

The embedded speech model fails to load STT models when the path contains non-ASCII characters and spaces.

To Reproduce

Steps to reproduce the behavior:

// the path below needs to exist on disk and needs to contain spaces and non-ASCII characters
auto speechConfig = EmbeddedSpeechConfig::FromPath("/Users/bpasero/Desktop/assets sind cööl/stt");

Expected behavior

Models can load from paths with spaces and non-ASCII characters.

Version of the Cognitive Services Speech SDK

1.35.0

Platform, Operating System, and Programming Language

• OS: macOS
• Hardware: ARM
• Programming language: C++

Additional context

Reported in https://github.com/microsoft/vscode/issues/206512

[pankopon]

This is generally the same with all SDK file access at the moment, including log files.

[github-actions[bot]]

This item has been open without activity for 19 days. Provide a comment on status and remove "update needed" label.

[bpasero]

This seems to be an acknowledged bug!

[pankopon]

@bpasero File paths with non-ASCII characters are not supported in the current SDK implementation, whether it's about embedded speech models, KWS models or log files. This is for future enchancement, no ETA yet.

[bpasero]

@pankopon understood. note that VS Code users can be impacted as soon as their usernames in the OS contains non-ASCII characters because we store VS Code extensions in the user-home directory, which typically includes the username verbatim. The Azure Speech model will be loaded directly from said extensions folder.

Would be great to see a fix for this eventually as there is no real workaround for people impacted 👍

[bpasero]

Hm, I wonder if a workaround could be to load the model with a relative file path? I have not tried that so far, but that could certainly help address this issue for our users.

[pankopon]

@bpasero It depends on what the working directory of the process that uses the Speech SDK is. The model path would need to be relative to this. If both the process working directory and the model location are under the user's home directory then a relative path could work.

[github-actions[bot]]

This item has been open without activity for 19 days. Provide a comment on status and remove "update needed" label.

[bpasero]

Please fix it :-)

[github-actions[bot]]

This item has been open without activity for 19 days. Provide a comment on status and remove "update needed" label.

[bpasero]

Still an issue.

[github-actions[bot]]

This item has been open without activity for 19 days. Provide a comment on status and remove "update needed" label.

[bpasero]

Still an issue.