Commit History

fix: restore 15s frame limit and 3-tier adaptive tiling for long movie sequential generation
a790c1f

Antigravity commited on

fix: pre-initialize image processor and restore screenplay GPU execution to avoid bnb cpu moves
b57dfaf

Antigravity commited on

fix: run screenplay generation on CPU to completely avoid GPU quota timeouts
b0c62a2

Antigravity commited on

chore: rename Space metadata and Gradio UI to CinemAI
23e251f

Antigravity commited on

fix(critical): reset frame cap to 97, fix continuity bug, remove subtitle
7be6d65

dimensionalpulsar commited on

fix: raise frame cap to 265 (~14.7s), 3-tier tiling, harden continuity chain
9958c96

dimensionalpulsar commited on

fix: split scenes into individual GPU calls to fix hang on videos >5s
f7bde67

dimensionalpulsar commited on

feat: batch all 8 scenes in a single GPU call and extract loopback last-to-first frame for absolute character continuity
a18889b

Antigravity commited on

fix: apply setpts=PTS-STARTPTS to force continuous browser-compatible playback timestamps
895d705

Antigravity commited on

feat: expose 8 individual scene download players in UI and normalize fps input rate for stable movie concat
1014c97

Antigravity commited on

fix: use filter_complex concat for perfect sequential frame normalization across scenes
ef8cd18

Antigravity commited on

fix: re-encode streams using libx264 and aac in concat to resolve playback issues
3026f59

Antigravity commited on

feat: optimize ZeroGPU file transfer with raw bytes and integrate gr.Progress() bar
9991eb2

Antigravity commited on

feat: implement LTX-2 movie generator with Gemma-3 screenplay director
8084f16

Antigravity commited on

fix: use original user audio directly in output video for accurate lip-sync (bypass audio VAE decoder); fix _write_audio mono-to-stereo auto-conversion to prevent ValueError on single-channel input
c8adf51

dimensionalpulsar commited on

fix: wrap flash_attn_interface import in try/except (OSError/ImportError) - prevents libcudart.so.12 crash at module load time on ZeroGPU; add graceful FA3 fallback to PyTorch SDPA in DEFAULT attention
2cae410

dimensionalpulsar commited on

fix: resolve 5 critical bugs - remove duplicate ModelLedger, pass gemma_path to pipeline, GPU duration=120, move text_encoder to cuda, fix transformers/torchaudio version pins
d45e5fd

dimensionalpulsar commited on