3 Commits

Author	SHA1	Message	Date
Ethanfel	c0b7ccbcee	fix: substitute empty_clip_feat for video features when no video present ... Zero features through bias-free Cond_MLP produce near-zero activations, not the learned null signal the model was trained with. Use empty_clip_feat (the learned null video embedding) just like empty_sync_feat for sync. Also improve text_prompt tooltip to encourage detailed CoT descriptions. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>	2026-03-27 22:13:22 +01:00
Ethanfel	45633788a4	debug: add latent and audio stats logging to T2A node ... Print fakes latent stats (mean/std/min/max) and audio pre-norm stats to diagnose whether diffusion output is numerically reasonable. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>	2026-03-27 22:06:39 +01:00
Ethanfel	618e7de64b	feat: PrismAudioTextOnly node with correct T5-Gemma encoding ... Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>	2026-03-27 18:09:11 +01:00