djdzzzz 9b5a94bb50 feat: H3 Prompt Planner + multi-material context analysis ...

- New H3 Prompt Planner (GGUF): video/image/audio context, H3-standard three-section output
- Three task modes: t2v, i2v, h3_multi_ref
- PyAV seek for fast video frame extraction
- Frontend panel with drag-drop upload, audio trimming, material preview
- Direct passthrough to official MiniMax H3 Reference to Video node
- ref_video_1/ref_image_1/ref_audio_1 prompt referencing
- Orphan llama-omni-server cleanup on startup
- VRAM cleanup: gc.collect() + cuda.empty_cache() after inference
- Updated README with H3 section, INSTALL.md, TUTORIAL.zh-CN.md

2026-08-03 13:57:44 +08:00

..

bernini_pe_worker.py

i2v: system-prompt format + generic FINAL regex with number prefix

2026-07-19 21:50:25 +08:00

h3_worker.py

feat: H3 Prompt Planner + multi-material context analysis

2026-08-03 13:57:44 +08:00

qwen35_worker.py

update README: Qwen3.5 node, thinking toggle, ref slots

2026-07-20 21:30:42 +08:00