Replying to @bdsqlsz @IamEmily2050
Doesn't seem terrible from a transformers update this morning. Claims it's out Jan 10.
16
Was a hint an audio model by QWen was forthcoming by @JustinLin610 in late October, but sadly haven't seen any other mentions of this before or since. (Perplexityed/Groked even)
Can still cross our fingers for early 2026
1
47
Replying to @oyacaro @wildmindai
These are from the 20b - takes forever even on top HW w/ the distilled models, but people saying the quality isn't good haven't used it.
2
19
Replying to @virtualfilmer @wildmindai
What? It's extremely heavy to run, but if you think the quality is bad you're not running the model properly. Most discussion is on discord in Banodoco.
Ex:
1
14
Replying to @tomfgoodwin
Had similar experience w Claude Opus in early September, "Yes you caught me"
1
83
Replying to @DMBTrivia @hellorob
Bonus: Claude Opus coding a node to do the opposite (adding a SORA2 watermark to something else / ie: fake "that's AI!!")
99
Was excited for a second, but this is still not a true reality: (Re: Perplexity research now / looked into this ~10months ago)
55
Replying to @martinleblanc
"The Soap Opera Effect" - So many people have their TVs set to interpolate up to 60fps and don't realize how badly that affects perception of TV and Movie viewing. Hollywood did PSAs: youtube.com/watch?v=1J0Dan0W…
2
271
Replying to @DMBTrivia @SolidEvidence
Had @perplexity_ai confirm that NV research has been funded almost exclusively by those w/ interests in NV.
47












