Voice

This key is used for persistent keypair signatures in the DAVE protocol. For further details, see the whitepaper.

Generating a keypair and signature can be done with the following Python pseudocode:

import base64
import requests
from cryptography.hazmat.primitives.asymmetric import ec
from cryptography.hazmat.primitives import hashes
from cryptography.hazmat.primitives import serialization

API_ENDPOINT = 'https://discord.com/api/v9'
private_key = ec.generate_private_key(ec.SECP256R1())
public_key = private_key.public_key().public_bytes(serialization.Encoding.X962, serialization.PublicFormat.CompressedPoint)

# https://datatracker.ietf.org/doc/html/rfc9000#name-variable-length-integer-enc
def encode_quic_varint(data):
    length = len(data)
    if length < 2**6:
        return bytes([length]) + data
    elif length < 2**14:
        return bytes([(1 << 6) | (length >> 8), length & 0xFF]) + data
    elif length < 2**30:
        return bytes([(2 << 6) | (length >> 24), (length >> 16) & 0xFF, (length >> 8) & 0xFF, length & 0xFF]) + data
    else:
        return bytes([(3 << 6) | (length >> 56), (length >> 48) & 0xFF, (length >> 40) & 0xFF, (length >> 32) & 0xFF, (length >> 24) & 0xFF, (length >> 16) & 0xFF, (length >> 8) & 0xFF, length & 0xFF]) + data

# https://datatracker.ietf.org/doc/html/rfc9420/#name-signing
def sign_with_label(private_key, label, content) -> bytes:
    label_bytes = b'MLS 1.0 ' + label.encode('ascii')
    sign_content = encode_quic_varint(label_bytes) + encode_quic_varint(content)
    signature = private_key.sign(sign_content, ec.ECDSA(hashes.SHA256()))
    return signature

# The session ID is the static_client_session_id value from the READY payload
session_id = '00000000-0000-0000-0000-000000000000'.encode('ascii')
signature = sign_with_label(private_key, "DiscordSelfSignature", session_id + b':' + public_key)

data = {
    'key_version': 1,
    'public_key': 'data:application/octet-stream;base64,' + base64.b64encode(public_key).decode('utf-8'),
    'signature': 'data:application/octet-stream;base64,' + base64.b64encode(signature).decode('utf-8'),
}
headers = {
    'authorization': 'token',
    # Rest of headers here
}

r = requests.put('%s/voice/public-keys' % API_ENDPOINT, json=data, headers=headers)
r.raise_for_status()

This key is used by another user to encrypt voice data. For further details, see the whitepaper.

There are currently several caveats for this endpoint:

• channel_id must point to a stage channel
• Current user must already have joined channel_id
• You must have the MUTE_MEMBERS permission to unsuppress yourself; you can always suppress yourself
• You must have the REQUEST_TO_SPEAK permission to request to speak; you can always clear your own request to speak
• You can only set request_to_speak_timestamp to the present or a future time

There are currently several caveats for this endpoint:

• channel_id must point to a stage channel
• Target user must already have joined channel_id
• You must have the MUTE_MEMBERS permission
• When unsuppressed, user accounts will have their request_to_speak_timestamp set to the current time; bot users will not
• When suppressed, the user will have their request_to_speak_timestamp removed