Drop mamba-ssm / causal-conv1d from requirements

Long-form chunking + RE-USE on reference

use gradio server (#2)

Drop xformers: FA3 Hopper kernel crashes on ZeroGPU H200

Bump torch 2.4.0 -> 2.8.0 for ZeroGPU compat

Pin pydantic==2.10.6 — root cause of the recurring bool-iter crash

Bump to gradio 5.7.1; drop pydantic/starlette/fastapi 4.x compat pins

Pin starlette<1.0 + fastapi<0.116 for gradio 4.44.1 TemplateResponse compat

Pin pydantic==2.10.6 to fix gradio_client schema bool TypeError

Pin SDK to 4.44.1 + monkey-patch gradio_client schema bool TypeError

Bump gradio to 5.x to fix gradio_client schema TypeError on bool

Pin torch/torchaudio==2.4.0 to fix ZeroGPU libcudart.so.13 mismatch

Refactor for ZeroGPU: lazy TTSServer load + @spaces.GPU decorator

Pin huggingface_hub<1.0 (gradio 4.44.1 still imports HfFolder)

DramaBox Space — initial app + vendored ltx2