It would be really awesome if a simple rebuild of the package could be done where the build command is ran without development flags so that this is actually usable in a server application. If I want to use this I basically have to take the output, and post it into my own private repo and remove PrintOutput function, and other logging nightmares so its usable. Luckily the dist is pretty small. I won't explain how you can use this in a server, just know that the other thing is pretty annoying for logging.

Hello! I'm wanting to explore the options of making SAM sing for a project. I'm wondering what the pitch system is based on. Is there any relative connection to the 12 TET system or is there any kind of chart that showcases the frequencies? Would love to learn more about it.

A server returning the contents from sam.buf8(...) doesn't automatically plays on the client.

It would be nice if there was a sam.toWavBuffer() function for this and similar use cases.

The code for it already exists in the RenderBuffer function found here, so an implementation of it could just be a refactor of that function:

const toWavBuffer = (audiobuffer: Uint8Array) => {
    // This function is basically a refactor of RenderBuffer

    const text2Uint8Array = (text: string) => {
        const buffer = new Uint8Array(text.length);
        text.split("").forEach((e, index) => {
            buffer[index] = e.charCodeAt(0);
        });
        return buffer;
    };

    // Calculate buffer size.
    const realbuffer = new Uint8Array(
            // ...
    );

    let pos = 0;
    const write = (buffer: Uint8Array) => {
        realbuffer.set(buffer, pos);
        pos += buffer.length;
    };

    //RIFF header
    write(text2Uint8Array("RIFF")); // chunkID
    write(Uint32ToUint8Array(audiobuffer.length + 12 + 16 + 8 - 8)); // ChunkSize
    write(text2Uint8Array("WAVE")); // riffType
    //format chunk
    write(text2Uint8Array("fmt "));
    write(Uint32ToUint8Array(16)); // ChunkSize
    write(Uint16ToUint8Array(1)); // wFormatTag - 1 = PCM
    write(Uint16ToUint8Array(1)); // channels
    write(Uint32ToUint8Array(22050)); // samplerate
    write(Uint32ToUint8Array(22050)); // bytes/second
    write(Uint16ToUint8Array(1)); // blockalign
    write(Uint16ToUint8Array(8)); // bits per sample
    //data chunk
    write(text2Uint8Array("data"));
    write(Uint32ToUint8Array(audiobuffer.length)); // buffer length
    write(audiobuffer);

    return realbuffer;
}

const DownloadBuffer = (audiobuffer: Uint8Array) => {
     const blob = new Blob([toWavBuffer(realbuffer)], { type: "audio/wav" });
     // Other code to download buffer
}

any thoughts on it? I understand if you do not want to expose that on the API for simplicity.

I would be more than glad to open a PR.

Also thank you so much for making this library 🙏

It appears that the exported wave files (using this.download function) are only 8-bit. They sound much worse than the spoken in-browser output. I messed around in the code for a bit modifying both the wave file header and the this.buf8 and this.buf32 functions and I couldn't find a way to make it sound better.

"Dipthong" is actually spelled "diphthong". The typo is all over the code.

When I have a sentence sometimes no matter how many times I press it won’t work.

In the comments, it's noted that the character '*' in the second position of a phoneme, such as 'R*', is a "wildcard".

This is a bit misleading, as "wildcard" would indicate that "S*" should match "SH". This isn't the case.

The '*' in the second character position indicates the phoneme is only a single character long.

When matching phonemes, SAM first checks the two characters at the current buffer position against the 2 character phones phonemes (that is, they don't have a '*' in the second character). If this fails SAM then checks the single character at the buffer position against all the single character phonemes (the ones that do have a '*' in the second position).

When certain strings are inputted an infinite loop happens where nothing will be spoken, the window will hang, and the developer console will spam output. Tested on the most current version of both Firefox and Chrome.

Steps to replicate:

1. Go to https://discordier.github.io/sam/
2. Type SoftwareSoftwareSoftwareSoftware in the text box
3. Press "Say" button

Other strings that produce same results:
dhaaaxaxaxaxaxaxaxaxaxaxaxaxax
CurrentlyCurrentlyCurrentlyCurrently

It seems to have to do with the repetition of characters with no spaces. If you add spaces in between the words it works fine. (ex: Currently Currently Currently Currently)

Here is a sample of the console output:

i dont know how to download audio.

Шаг 1 вы с другом в заброшенной психиатрической больнице

I'm not seeing a way

took me about 15 mins before i finally figured out that libsdl1.2-dev was the package required to compile the code and get it to work.
even if this doesn't get fixed those who are gonna do this will atleast know what they are missing instead of getting confused.

Import doesn't work.



Help plz. I'm a complete zero in js.

Hard for me to verify the legitimacy of this but it appears at at some point someone reached out to Joseph Katz regarding either this project or a similar one. I can't find contact info for Joseph but it appears he is willing to grant individuals permission to use the software for non-commercial use

. I've been doing heavy research into tracking him down as I wanted to use it in something commercial but I'm not sure what the answer would be on that.

In any case it might be worth looking into.

https://nvda.groups.io/g/nvda/topic/softvoice_information_request/5170758

I copied the code, yet it says: "[error] Line 1: Uncaught SyntaxError: Unexpected identifier 'add'".
Is there something missing or this code isn't for JavaScript?

The original C version of SAM is pretty buggy - this one seems to be far better maintained, but it's in JavaScript. Any idea how I can easily embed this in a C++ (MSVC) program?

I need this for one of my projects. I will be very grateful if you help!

this isn't that bad but when using longer messages then changing them, you'd think hitting "say" again would just restart it but no, it causes it to start saying the beginning and where it was at the same time creating a jumbled mess (of course refreshing the page fixes this but I just wanted to point it out)

Hello!

I'd like to use the SAM in one WEB-game, so I have question. Did I understand correctly that I can use samjs.js-file as a standalone script? Do I need any other scripts for working SAM in browser? I've experimented in Chrome, TTS is working with this one file, but maybe I don't know something...

And one another question, how I can enable the phonetics mode?

Thanks

The table frequencyData values are referred to in the comments frequency values, which isn't strictly true.

They're actually stride values. Per your comments in the rendering sections, the original version of SAM used a pre-calculated sine table instead of calling the sin() function.

By using a stride value, SAM is able to find the next sine value for each formant simply by adding the stride value to the table index (and, of course, wrapping at the end of the table).