I wanted to find more "heavy metal with trance-like synthesizers" and didn't really get any good results. To give the perfect example of what I'm looking for, see Blood Stain Child - Stargazer: https://youtu.be/IjTXeJZM8wg
DALL-E for Playlists
61–70 of 99 posts
Re: DALL-E for Playlists
#62Gives you an insight into the hilarious prompts others here are using
Re: DALL-E for Playlists
#63Track 8: Sandwiches
Re: DALL-E for Playlists
#64I highly recommend everyone look through the other playlists on Spotify that the account also generates. Gives you an insight into the hilarious prompts others here are using
Re: DALL-E for Playlists
#65OP HERE! Solved the hugging to death issue. Reached the Spotify 11k playlist limit. Deleting old playlists and the site should be working as intended now. THANK YOU whoever posted this! Currently working on: Fixing the recommendation scoring function. Right now it's giving hit or miss responses. I think the problem is with my cross encoder "reranker" is not doing its job the right way. I'll fix the passages it looks…
I haven't tried it yet, but this is exactly how I've always wanted to have playlists generated, so I'm looking forward to trying it! Auto-generated playlists are sometimes great, but IMO they often miss the mark of what I wanted or why I like a particular song. So I think this kind of playlist generation could help solve this, especially if the playlist can be refined after initial generation.
Re: DALL-E for Playlists
#66I also tried "Cover songs which are much more famous than the originals", and after spot checking the list, it seems to be originals - not covers.
I gave the same prompts to ChatGPT. It wouldn't even try with the first prompt, but the second prompt yielded a decent list.
Re: DALL-E for Playlists
#67Wrote the prompt as a joke but the end result is not bad at all!
Re: DALL-E for Playlists
#68Earlier quoted context omitted.
TLDR: Lots of musical metadata converted into paragraphs and the SentenceTransformers Retrieve & Re-Rank Pipeline.( https://www.sbert.net/index.html ) The sentence embeddings are calculated using a Bidirectional Encoder Representation Transformer (BERT) model. There's a pre-trained model for this network trained on over 1 billion sentences from the internet that is publicly available, (thanks Microsoft) . The model t…
That's awesome, thank you so much for sharing all that. The project is great.
Re: DALL-E for Playlists
#69OP HERE! Solved the hugging to death issue. Reached the Spotify 11k playlist limit. Deleting old playlists and the site should be working as intended now. THANK YOU whoever posted this! Currently working on: Fixing the recommendation scoring function. Right now it's giving hit or miss responses. I think the problem is with my cross encoder "reranker" is not doing its job the right way. I'll fix the passages it looks…
Re: DALL-E for Playlists
#70I asked for a playlist of "how ADHD sounded like" and I can't be more impressed. I don't have ADHD myself but live with people that do and I think the playlist really captured some of the observed symptoms of the disorder in a musical way. But what I really found impressive is: searching for "ADHD playlists" in Spotify yields a lot of playlists with lo-fi music for deep focus, not the fast-paced, disorganized, hypera…
The themes I can detect in the playlist itself are "I don't know quite what this is, but that itself is interesting, so my brain is just going to surf on that for a while", "nothing's going on, so I started daydreaming and my brain is in that comfortable relaxed happy place", and "oh oh oh, Yes yes yes! Mode"