Play from youtube skill

pcwii · July 3, 2020, 5:51pm

I am having a similar issue with my Kodi-skill, it is no longer returning youtube links either so something has changed.

gnicolas31 · July 3, 2020, 7:49pm

Acutally trying to debug it (lol first time I’m seeing python script). It seem that if you change skill to load only one music it’s working. So I think that the issue is with youtube-dl who can scrap youtube video because maybe youtube has made some changes

pcwii · July 3, 2020, 8:07pm

Looks like the playlist ID is formatted differently. The regex is failing in my kodi-skill.
These no longer work for me…

all_video_links = re.findall(r'href=\"\/watch\?v=(.{11})', html.decode())
all_playlist_results = re.findall(r'href=\"\/playlist\?list\=(.{34})', html.decode())

These do work…

all_video_links = re.findall(r'/watch\?v=(.{11})', html.decode())
all_playlist_results = re.findall(r'list\=(.{18})', html.decode())

Not sure this is any help to anyone but me but thought I would put it out there. I am going to dig in when I get some time.

gnicolas31 · July 4, 2020, 1:43am

I tried to download Jarbais plugin but the github seem no longer maintained, do you have a fork ?

antdog_evervast · July 8, 2020, 6:21pm

Anyone having any luck with youtube… having same issue with youtube skill stopped working few days ago … saying “no results found”

gnicolas31 · July 8, 2020, 6:24pm

Same issue here. Was using Mcdruid plugin and since few days it’s ‘no result found’

pcwii · July 8, 2020, 7:01pm

There are several youtube skills scattered throughout the community. What skill(s) are having issues. From my own testing it looks like the regex is failing on most of these due to a change in youtube. I have seen a similar issue with my Kodi-skill that I am correcting.

@mcdruid may be able to assist with their skill.

antdog_evervast · July 8, 2020, 7:18pm

this is what i found in @mcdruid yoiutube skill

for vid in vids:
if not re.match(’/watch?v=\w{11}’, vid[‘href’]):
LOG.debug('no media: ’ + vid[‘href’])
continue

antdog_evervast · July 8, 2020, 7:19pm

the whole section

Attempt to find the first result matching the query string

def search_youtube(self, search_term):
    tracklist = []
    res = requests.get(search_url + search_term)
    # TODO: check status code etc...
    html = res.content
    soup = BeautifulSoup(html, 'html.parser')
    vids = soup.findAll(attrs={'class':'yt-uix-tile-link'})

    for vid in vids:
        if not re.match('/watch\?v=\w{11}', vid['href']):
          LOG.debug('no media: ' + vid['href'])
          continue

        self.vid_url = vid['href']
        self.vid_name = vid.string
        self.stream_url = self.get_stream_url(self.vid_url)
        LOG.debug('Found stream URL: ' + self.vid_url)
        LOG.debug('Media title: ' + self.vid_name)
        tracklist.append(self.stream_url)
        self.mediaplayer.add_list(tracklist)
        self.audio_state = 'playing'
        self.speak_dialog('now.playing', {'content': self.vid_name} )
        wait_while_speaking()
        self.mediaplayer.play()
        return

antdog_evervast · July 8, 2020, 7:42pm

my bots arnt singing to me while i work on them…lol…

pcwii · July 8, 2020, 7:57pm

issue is with this line…

vids = soup.findAll(attrs={'class':'yt-uix-tile-link'})

This class no longer exists.

pcwii · July 8, 2020, 8:04pm

@mcdruid has been MIA for a bit. Not sure if he will respond or not. If I get adventurous maybe I will see if I can get his skill working. I have an idea what needs to be done.

mcdruid · July 8, 2020, 9:42pm

It looks like youtube have completely changed the way they’re returning results.

Searching the HTML for the yt-uix-tile-link class doesn’t look like it’s going to work any more.

As far as I can see, they may now be returning a chunk of json which is rendered client side.

So we may need to look at parsing that instead.

Or, perhaps you can send a different User Agent and get plain HTML back… not sure.

Parsing the json is probably a better approach anyway.

I’ll have a look at this but don’t have a lot of time to work on it at the moment.

Pull Requests welcome at https://gitlab.com/mcdruid/mycroft-youtube-audio

pcwii · July 8, 2020, 10:13pm

@mcdruid, Glad to see you are still active, I don’t have a gitlab account but have been doing a bit of playing around. There may be an issue with yt-downloader too but I am unsure since I think i have maxed out my google quota with testing I have been playing with your code and changed your search_youtube routine to this code and it seams to return valid results.


    def search_youtube(self, search_term):
        tracklist = []
        res = requests.get(search_url + search_term)
        # TODO: check status code etc...
        html = res.content
        soup = BeautifulSoup(html, 'html.parser')
        all_vid_id = re.findall(r'/watch\?v=(.{11})', str(soup))
        if len(all_vid_id) >= 1:
            for vid_id in all_vid_id:
                vid_url = "/watch?v=" + str(vid_id)
                self.stream_url = self.get_stream_url(vid_url)
                LOG.debug('Found stream URL: ' + self.stream_url)
                tracklist.append(self.stream_url)
            LOG.info(str(tracklist))
            self.mediaplayer.add_list(tracklist)
            vid_name = str(search_term)
            self.audio_state = 'playing'
            self.speak_dialog('now.playing', {'content': vid_name})
            wait_while_speaking()
            self.mediaplayer.play()
            return
        else:
            # We didn't find any playable results
            self.speak_dialog('not.found')
            wait_while_speaking()
            LOG.debug('Could not find any results with the query term: ' + search_term)

mcdruid · July 8, 2020, 10:27pm

Ok, great so that approach is just finding the watch?v=BLAH url in the soup, and using the original search term as the title.

Enough to get something work again, but it’s a shame not to be parsing the title from the result etc…

I’ve started to look at using something like https://stackoverflow.com/a/56059878 to extract and parse the json. Looks sort of okay, but with the first search terms I’ve tried it’s failing to parse the json properly, and I’m not sure why yet…

JarbasAl · July 9, 2020, 2:26am

you might want to look at

for those using the gui @Aix has done a good job

i think we should concentrate dev efforts into a single place, consider sending PRs to the more advanced skill instead

pcwii · July 9, 2020, 2:44am

I agree, there is definitely a lot of duplication and overlap when it comes to playing music. Not sure of the solution. I have also noticed many of the cps music skills return a match even before confirming they actually have the resource or music available. …Someone says “play xyz” and they return exact match before checking if xyz is playable or available to this skill.

gnicolas31 · July 10, 2020, 6:51pm

Guys, just for information, I’ve forked the Mcdruid Repo with the pcwii update if anyone need , available here :

antdog_evervast · July 10, 2020, 10:01pm

Awesome thank you… I would sing to my robots but I am trying to teach them to walk… not run away… lol

luispv · December 11, 2020, 1:50pm

Thanks, it worked for me!
Just two things:

I got the following error:
Preformatted text~~~~nnection object at 0x4405dff0>: Failed to establish a new connection: [Errno -5] No address associated with hostname’))
It takes too much to play the song

Again, thanks.