13 Commits

Author	SHA1	Message	Date
kijai	64a984177f	Update model_multitalk.py	2025-11-26 18:15:12 +02:00
kijai	af4d412e67	Chunk attention map calculation for multiple speakers to reduce peak VRAM usage	2025-11-26 18:14:54 +02:00
kijai	fb099a40b2	Handle ref_attn_mask with separate patch to avoid having to always return q and k from self_attn	2025-11-05 17:24:13 +02:00
kijai	3ae78a4804	Update model_multitalk.py	2025-11-03 21:13:57 +02:00
kijai	d53e62913d	Remove looping functionality, keep extension functionality	2025-11-03 21:12:02 +02:00
kijai	f5d53f2a6b	Restore preview functionality	2025-10-23 23:50:09 +03:00
kijai	7842a5c805	remove import	2025-10-10 21:45:10 +03:00
kijai	9c5022e7e3	this is redundant	2025-10-10 21:40:48 +03:00
kijai	57567bde4e	remove print	2025-10-03 18:28:46 +03:00
kijai	00c069dd1c	Update model_multitalk.py	2025-10-03 18:27:29 +03:00
kijai	6f6db12bbe	whitespace...	2025-10-03 18:26:46 +03:00
kijai	460ce7f77b	Update model_multitalk.py	2025-10-03 18:17:06 +03:00
kijai	efe83f5a36	re-init	2025-10-03 17:24:05 +03:00