Unable to process audio from stage channels | Pycord | Page 1

valid mesa Nov 16, 2024, 11:17 PM

#

When I attempt to process audio after my bot joins a stage channel, I get this error:

Traceback (most recent call last):
  File "/usr/lib/python3.8/threading.py", line 932, in _bootstrap_inner
    self.run()
  File "/usr/lib/python3.8/threading.py", line 870, in run
    self._target(*self._args, **self._kwargs)
  File "/home/ubuntu/project/lib/python3.8/site-packages/discord/voice_client.py", line 863, in recv_audio
    self.unpack_audio(data)
  File "/home/ubuntu/project/lib/python3.8/site-packages/discord/voice_client.py", line 740, in unpack_audio
    data = RawData(data, self)
  File "/home/ubuntu/project/lib/python3.8/site-packages/discord/sinks/core.py", line 115, in __init__
    self.decrypted_data = getattr(self.client, f"_decrypt_{self.client.mode}")(
  File "/home/ubuntu/project/lib/python3.8/site-packages/discord/voice_client.py", line 611, in _decrypt_xsalsa20_poly1305_lite
    return self.strip_header_ext(box.decrypt(bytes(data), bytes(nonce)))
  File "/home/ubuntu/project/lib/python3.8/site-packages/nacl/secret.py", line 149, in decrypt
    plaintext = nacl.bindings.crypto_secretbox_open(
  File "/home/ubuntu/project/lib/python3.8/site-packages/nacl/bindings/crypto_secretbox.py", line 79, in crypto_secretbox_open
    ensure(
  File "/home/ubuntu/project/lib/python3.8/site-packages/nacl/exceptions.py", line 88, in ensure
    raise raising(*args)
nacl.exceptions.CryptoError: Decryption failed. Ciphertext failed verification

I see an open but unresolved github issue about it: https://github.com/Pycord-Development/pycord/issues/2033.

I tested the same code when joining a regular voice channel, and it works. Just a stage channel. I've also experimented with making the bot a speaker, muting, etc. Nothing seems to work.

Note: I've tried upgrading python version and pycord version to 2.6.1, but it doesn't change the error.

GitHub

Issues · Pycord-Development/pycord

Pycord is a modern, easy to use, feature-rich, and async ready API wrapper for Discord written in Python - Issues · Pycord-Development/pycord

tight haven Nov 16, 2024, 11:19 PM

#

valid mesa When I attempt to process audio after my bot joins a stage channel, I get this e...

Install Python 3.9

valid mesa Nov 16, 2024, 11:19 PM

#

tight haven Install Python 3.9

I installed python 3.10 and the exact same error happens

tight haven Nov 16, 2024, 11:19 PM

#

The support for 3.8 got dropped

valid mesa Nov 16, 2024, 11:19 PM

#

It's unrelated

tight haven Nov 16, 2024, 11:19 PM

#

the Error you send it 3.8

valid mesa Nov 16, 2024, 11:20 PM

#

tight haven the Error you send it 3.8

Yeah I can paste the logs again it'll say the exact same thing only with v3.10. This is just from an older run where I saved it

#

Seems to be an open issue so I'm not sure why this is but looks like it is isolated to stage channels

unborn herald Nov 16, 2024, 11:21 PM

#

valid mesa It's unrelated

i told you over a week or what ago that 3.8 is outdated. you should update no matter if its related or not

valid mesa Nov 16, 2024, 11:21 PM

#

unborn herald i told you over a week or what ago that 3.8 is outdated. you should update no ma...

I updated it. Please read the full post

#

I get the exact same error

unborn herald Nov 16, 2024, 11:21 PM

#

"no matter if its related or not"

#

read my message too

valid mesa Nov 16, 2024, 11:21 PM

#

I'm using 3.10 now

unborn herald Nov 16, 2024, 11:22 PM

#

which is also 3 years old by now

valid mesa Nov 16, 2024, 11:22 PM

#

what version should I use

scenic aurora Nov 16, 2024, 11:22 PM

#

at this point it is not a python version problem. And pycord does support 3.10 so the bug should be fixed regardless

valid mesa Nov 16, 2024, 11:23 PM

#

If anyone has any suggestions would greatly appreciate it

#

It's weird that it works on regular voice channels

unborn herald Nov 16, 2024, 11:23 PM

#

valid mesa I'm using 3.10 now

ideally 3.13 because thats the newest version
ill never understand why people run 6+ year old versions of python

scenic aurora Nov 16, 2024, 11:23 PM

#

Does this happen immediately or only after a long time of listening like in the github issue?

unborn herald Nov 16, 2024, 11:23 PM

#

but 3.12 currently because theres a bug with pycord on 3.13

valid mesa Nov 16, 2024, 11:23 PM

#

scenic aurora Does this happen immediately or only after a long time of listening like in the ...

No actually, it happens right away

#

that's the other odd thing

#

It errors out immediately, if there's sound playing or even if there's nothing playing

unborn herald Nov 16, 2024, 11:24 PM

#

im not sure if this is needed anymore but did you install pycord[voice] ?

valid mesa Nov 16, 2024, 11:24 PM

#

No I didn't. I installed pycord=2.6.1

unborn herald Nov 16, 2024, 11:24 PM

#

try pycord[voice]

valid mesa Nov 16, 2024, 11:26 PM

#

unborn herald try pycord[voice]

no dice, same issue

#

Does this have anything to do with it? #discussion message

unborn herald Nov 16, 2024, 11:30 PM

#

well its a nacl issue from what i can see

#

or at least nacl throws the error

valid mesa Nov 16, 2024, 11:31 PM

#

valid mesa Does this have anything to do with it? https://discord.com/channels/881207955029...

This message references xsalsa20_poly1305 which I see in the logs of the error

unborn herald Nov 16, 2024, 11:33 PM

#

well technically its only deprecated until tomorrow lol

scenic aurora Nov 16, 2024, 11:33 PM

#

Thats what I am wondering, we nelo said we are fine but I wonder if that is wrong.
In any case that should still work until the 18th though.

unborn herald Nov 16, 2024, 11:34 PM

#

or discord just said fuck it and removed it earlier

#

would fit discord

#

on an unrelated note i wonder why almost every python library targets voice API version 4
like thats so old, and they all do it, but only for python

#

surely just targeting the newest one would prevent such issues

#

my only idea would be to manually update the nacl package to the newest version

valid mesa Nov 16, 2024, 11:39 PM

#

unborn herald my only idea would be to manually update the nacl package to the newest version

I'll give this a try

scenic aurora Nov 16, 2024, 11:40 PM

#

unborn herald on an unrelated note i wonder why almost every python library targets voice API ...

Probably because no one wants to touch voice. We dont really even have anyone here who is knowledgeable at voice to my knowledge.

unborn herald Nov 16, 2024, 11:40 PM

#

lol
so whats the plan for tomorrow when its officially supposed to be removed?

#

like the 2 algorithms or whatever

scenic aurora Nov 16, 2024, 11:41 PM

#

Nelo said it was fine because 4 is still supported. I dont think it is tho...

unborn herald Nov 16, 2024, 11:41 PM

#

it is according to that message

#

just not those 2 whatever-it-is

scenic aurora Nov 16, 2024, 11:46 PM

#

So from what I can tell it seems like the data is somehow malformed/changed so nacl cannot decrypt it.
The question is if discord is being weird or if pycord reads it wrong when receiving or something

valid mesa Nov 16, 2024, 11:49 PM

#

scenic aurora So from what I can tell it seems like the data is somehow malformed/changed so n...

It's happening right away, so I'm going to try adding some sleep and maybe ignoring the initial data packets. I'll see if that works

scenic aurora Nov 16, 2024, 11:50 PM

#

https://github.com/Pycord-Development/pycord/blob/e4946ecf5c52a1c2900beba45281b6d052444383/discord/sinks/core.py#L111
This seems strange, why are we chopping off the first 12 bytes? (bits?)

unborn herald Nov 16, 2024, 11:53 PM

#

why do people not comment when they do seemingly nonsenical stuff like that

#

a simple "cutting off 12 bytes cause" wouldve helped already

scenic aurora Nov 16, 2024, 11:54 PM

#

The deprecated encryption modes above will be discontinued as of November 18th, 2024. As of this date the voice gateway will not allow you to connect with one of the deprecated encryption modes.
-# https://discord.com/developers/docs/topics/voice-connections#transport-encryption-modes
I think this is the issue

unborn herald Nov 16, 2024, 11:54 PM

#

yea but its not nov 18 in any country

scenic aurora Nov 16, 2024, 11:55 PM

#

they are already deprecated according to this and will be discontinued on Nov 18th. Unless that is a typo

unborn herald Nov 16, 2024, 11:55 PM

#

yea but deprecated means it still works

scenic aurora Nov 16, 2024, 11:56 PM

#

Oh, i read that wrong, they worded it weird.

#

nevermind

#

Ok, the 12 bytes being chopped off are because of this https://discord.com/developers/docs/topics/voice-connections#transport-encryption-modes-voice-packet-structure

scenic aurora Nov 17, 2024, 12:28 AM

#

To be honest. Everything looks fine. The only thing I can think of is that discord is sending a voice packet of more than 4096 bytes. I dont know if the max size is documented anywhere

scenic aurora Nov 17, 2024, 1:03 AM

#

@valid mesa
Can you try putting this code before you define the voice client object in your code

VoiceClient.supported_modes = (
        "xsalsa20_poly1305",
    )

If that does not work try

VoiceClient.supported_modes = (
        "xsalsa20_poly1305_suffix",
    )

valid mesa Nov 17, 2024, 1:04 AM

#

scenic aurora <@710614833602625618> Can you try putting this code before you define the voice...

Let me try this, thanks

scenic aurora Nov 17, 2024, 1:06 AM

#

Can you show me that section of the code. Just so I know you are doing it right (I dont want us to get confused

valid mesa Nov 17, 2024, 1:08 AM

#

scenic aurora Can you show me that section of the code. Just so I know you are doing it right ...

    @tasks.loop(seconds=10.0)  # Reduced interval for quicker detection
    async def search_for_live(self) -> None:
        if not self.is_in_stage:
            voice_client = await could_join_stage(self.bot)
            if voice_client:
                self.is_in_stage = True
                self.voice_client = voice_client
                self.voice_client.supported_modes = ("xsalsa20_poly1305",)
                await self.send_log_message(message="Joined stage!")
                await self.start_recording()

voice_client comes from the could_join_stage() function via this line:

...
voice_client = await channel.connect()
return voice_client

valid mesa Nov 17, 2024, 1:14 AM

#

scenic aurora <@710614833602625618> Can you try putting this code before you define the voice...

No luck =/

scenic aurora Nov 17, 2024, 1:15 AM

#

I think you need to do it on the actuall VoiceClient class not an object

#

I think you could just import the class in your main file or something and just paste the 2 snippets I sent above in there one at a time

valid mesa Nov 17, 2024, 1:17 AM

#

scenic aurora I think you could just import the class in your main file or something and just...

let me try

valid mesa Nov 17, 2024, 1:24 AM

#

scenic aurora I think you could just import the class in your main file or something and just...

For the second method now I get:

TypeError: initializer for ctype 'unsigned char *' must be a cdata pointer, not bytearray

Looked at the trace is still coming from nacl

scenic aurora Nov 17, 2024, 1:45 AM

#

valid mesa For the second method now I get: ``` TypeError: initializer for ctype 'unsigned ...

Could you paste the full traceback. I think this is going backwards though not forwards. I really dont know what to tell you.
To be fair I think all of voice is going to break on the 18th so there might be a whole new implementation anyways

valid mesa Nov 17, 2024, 1:52 AM

#

scenic aurora Could you paste the full traceback. I think this is going backwards though not f...

File "/home/ubuntu/project/lib/python3.10/site-packages/discord/voice_client.py", line 863, in recv_audio
    self.unpack_audio(data)
  File "/home/ubuntu/project/lib/python3.10/site-packages/discord/voice_client.py", line 740, in unpack_audio
    data = RawData(data, self)
  File "/home/ubuntu/project/lib/python3.10/site-packages/discord/sinks/core.py", line 115, in __init__
    self.decrypted_data = getattr(self.client, f"_decrypt_{self.client.mode}")(
  File "/home/ubuntu/project/lib/python3.10/site-packages/discord/voice_client.py", line 602, in _decrypt_xsalsa20_poly1305_suffix
    return self.strip_header_ext(box.decrypt(bytes(data[:-nonce_size]), nonce))
  File "/home/ubuntu/project/lib/python3.10/site-packages/nacl/secret.py", line 149, in decrypt
    plaintext = nacl.bindings.crypto_secretbox_open(
  File "/home/ubuntu/project/lib/python3.10/site-packages/nacl/bindings/crypto_secretbox.py", line 78, in crypto_secretbox_open
    res = lib.crypto_secretbox_open(plaintext, padded, len(padded), nonce, key)
TypeError: initializer for ctype 'unsigned char *' must be a cdata pointer, not bytearray

#

I don't know what type of data it's trying to pass here. Ignoring the exception doesn't seem to work but I'm still messing with it

scenic aurora Nov 17, 2024, 2:07 AM

#

What is the output of

import nacl
print(nacl.secret.SecretBox.NONCE_SIZE)

valid mesa Nov 17, 2024, 2:11 AM

#

>>> import nacl
>>> print(nacl.secret.SecretBox.NONCE_SIZE)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
AttributeError: module 'nacl' has no attribute 'secret'

#

hmmmm

#

brb

scenic aurora Nov 17, 2024, 2:12 AM

#

uhhh, try import nacl.secret?

scenic aurora Nov 17, 2024, 7:46 PM

#

@valid mesa You can try using the PR for the new voice encodings.
pip install git+https://github.com/Pycord-Development/pycord.git@refs/pull/2651/head
This is going to be required tomorrow so no point trying to debug the old ones at this point

remote python Jan 18, 2025, 1:15 PM

#

valid mesa ``` >>> import nacl >>> print(nacl.secret.SecretBox.NONCE_SIZE) Traceback (most ...

Did you solve the issue?

arctic pier Jan 18, 2025, 9:08 PM

#

remote python Did you solve the issue?

This is going to be required tomorrow so no point trying to debug the old ones at this point

remote python Jan 18, 2025, 9:08 PM

#

arctic pier > This is going to be required tomorrow so no point trying to debug the old ones...

wdym?

arctic pier Jan 18, 2025, 9:09 PM

#

remote python wdym?

Voice encoding changed. There's no point on debugging a bug on the old encodings methods.

arctic pier Jan 18, 2025, 9:09 PM

#

scenic aurora <@710614833602625618> You can try using the PR for the new voice encodings. `pip...

Literally this.

scenic aurora Jan 20, 2025, 3:39 PM

#

@remote python I think I fixed it and you can make the changes manually based on the diff here https://github.com/NeloBlivion/pycord/pull/2
The problem seemed to be that discord was sending non-opus packets that we need to ignore.

remote python Jan 20, 2025, 5:14 PM

#

DaniSalute

scenic aurora Jan 20, 2025, 5:16 PM

#

Huh, I completely selected the wrong thread when sending this message.

remote python Jan 20, 2025, 5:16 PM

#

all good

#

saw the message

#Unable to process audio from stage channels