Hi, so I get this link error LINK : fatal error LNK1181: cannot open input file 'stt.lib' so I guess I need to download this lib right and point to it somehow maybe ? yes I am new to this speech tech :) and then i should have a pretrained model right? can you please give me a step by step for speech to text using this thank you

This PR goes in tandem with this one in coqui-stt-sys to fix the same issues.

In that PR, new link attributes were added, so adding them manually in build.rs is not needed anymore.

Also, instead of adding C:\stt to the library search paths manually in build.rs, the same effect can be achieved by adding a library folder to the PATH env variable, which is the common solution to having global libraries on windows. If the developer doesn't want to change path, they can always create their own build.rs and add their own library search directory.

Lastly, as mentioned in the coqui-stt-sys PR, some bindings are not available on windows, so their high-level equivalents should also be removed.

It looks like libcoqui-sys does not automatically grab the C library, so it would be nice to document in the README.md how and where to get it, and which version this will work with.

Hi there @tazz4843 👋

forgive my Rust ignorance:)

I was hoping to have an easy install with cargo install, but I get this error:

$ cargo install coqui-stt
    Updating crates.io index
  Downloaded coqui-stt v0.2.1
  Downloaded 1 crate (15.0 KB) in 1.74s
error: there is nothing to install in `coqui-stt v0.2.1`, because it has no binaries
`cargo install` is only for installing programs, and can't be used with libraries.
To use a library crate, add it as a dependency in a Cargo project instead.

I see what the error message, and what little I've read says this could be used as a dependency.

If it would make sense, I think it would be a great dev experience to install via a package manager like cargo

thanks for open-sourcing this btw!

Hey, thanks for this. May we have more examples? Particularly, streaming audio.

I can't find the threadsafe-streams feature in the crate and I want to use intermediate decode. Basically, I want to do multiple things with my audio stream. With coqui, I want to spawn a thread to do STT.

        stream.play().unwrap(); //from cpal
        streaming.feed_audio(&audio_buf_stream);
        println!("feeding audio");
        while let Ok(o) = streaming.intermediate_decode() {
            println!("{o}");
        }

I'm just getting eeeee as continuous output. Which also makes me wonder if I can transfer deepspeech models to here? Or would I have to retrain for coqui specifically since i'm getting garbage output? It's a tflite model.

I'm all over the place, but I'm just asking for an example to do speech to text in a new thread. And how would I enable threadsafe-streams feature, I can't find it

I've read in other issues that you tried to get static compilation, but in your read me you just say "static compilation is impossible". Why is it impossible?

I noticed that Send and Sync impls were removed from Stream, with the comment that they were unsafe, however looking at the source code for libstt it doesn't seem to be doing anything that would prevent them from being Send or Sync. The closest I could find was a thread in the discussions for STT about a segfault another developer was having, but they mentioned using a mutex to fix the issue on Android, so this seems like more of an issue of improper management of shared mutable state than a fundamental lack of thread safety in Stream. Is there another reason why Stream can't be shared between threads even with proper synchronization?

coqui-stt could not be compiled for the target armv7-unknown-linux-gnueabihf because STT_CreateModelFromBuffer requires an argument whose type is std::os::raw::c_char which is not i8 on the previously mentioned target. Casting the argument to std::os::raw::c_char before passing it will make the library more os-agnostic.