+        return formats
+
+    def _real_extract(self, url):
+        mobj = re.match(self._VALID_URL, url)
+        domain_id = mobj.group('domain_id') or mobj.group('domain_id_s')
+        video_id = mobj.group('id')
+
+        video = None
+
+        def find_video(result):
+            if isinstance(result, dict):
+                return result
+            elif isinstance(result, list):
+                vid = int(video_id)
+                for v in result:
+                    if try_get(v, lambda x: x['general']['ID'], int) == vid:
+                        return v
+            return None
+
+        response = self._download_json(
+            'https://arc.nexx.cloud/api/video/%s.json' % video_id,
+            video_id, fatal=False)
+        if response and isinstance(response, dict):
+            result = response.get('result')
+            if result:
+                video = find_video(result)
+
+        # not all videos work via arc, e.g. nexx:741:1269984
+        if not video:
+            # Reverse engineered from JS code (see getDeviceID function)
+            device_id = '%d:%d:%d%d' % (
+                random.randint(1, 4), int(time.time()),
+                random.randint(1e4, 99999), random.randint(1, 9))
+
+            result = self._call_api(domain_id, 'session/init', video_id, data={
+                'nxp_devh': device_id,
+                'nxp_userh': '',
+                'precid': '0',
+                'playlicense': '0',
+                'screenx': '1920',
+                'screeny': '1080',
+                'playerversion': '6.0.00',
+                'gateway': 'html5',
+                'adGateway': '',
+                'explicitlanguage': 'en-US',
+                'addTextTemplates': '1',
+                'addDomainData': '1',
+                'addAdModel': '1',
+            }, headers={
+                'X-Request-Enable-Auth-Fallback': '1',
+            })
+
+            cid = result['general']['cid']
+
+            # As described in [1] X-Request-Token generation algorithm is
+            # as follows:
+            #   md5( operation + domain_id + domain_secret )
+            # where domain_secret is a static value that will be given by nexx.tv
+            # as per [1]. Here is how this "secret" is generated (reversed
+            # from _play.api.init function, search for clienttoken). So it's
+            # actually not static and not that much of a secret.
+            # 1. https://nexxtvstorage.blob.core.windows.net/files/201610/27.pdf
+            secret = result['device']['clienttoken'][int(device_id[0]):]
+            secret = secret[0:len(secret) - int(device_id[-1])]
+
+            op = 'byid'
+
+            # Reversed from JS code for _play.api.call function (search for
+            # X-Request-Token)
+            request_token = hashlib.md5(
+                ''.join((op, domain_id, secret)).encode('utf-8')).hexdigest()
+
+            result = self._call_api(
+                domain_id, 'videos/%s/%s' % (op, video_id), video_id, data={
+                    'additionalfields': 'language,channel,actors,studio,licenseby,slug,subtitle,teaser,description',
+                    'addInteractionOptions': '1',
+                    'addStatusDetails': '1',
+                    'addStreamDetails': '1',
+                    'addCaptions': '1',
+                    'addScenes': '1',
+                    'addHotSpots': '1',
+                    'addBumpers': '1',
+                    'captionFormat': 'data',
+                }, headers={
+                    'X-Request-CID': cid,
+                    'X-Request-Token': request_token,
+                })
+            video = find_video(result)
+
+        general = video['general']
+        title = general['title']
+
+        cdn = video['streamdata']['cdnType']
+
+        if cdn == 'azure':
+            formats = self._extract_azure_formats(video, video_id)
+        elif cdn == 'free':
+            formats = self._extract_free_formats(video, video_id)
+        else:
+            # TODO: reverse more cdns
+            assert False
+