Spaces:
Sleeping
Sleeping
Commit History
-
Remove the LLM compose step (Inference Providers cost)
Lyrics + structured caption are now editable reference nodes wired directly into generate_song, with the official-guide defaults. Drops the openai dep, the MiniMax-M3 prompts, OAuth token plumbing, and the inference-api scope.cffc867 -
Note optional sign-in for custom lyrics in the workflow description
The canvas already provides the LoginButton; no extra button needed.4c1e11d -
Convert to gr.Workflow canvas app on ZeroGPU
- app.py: gr.Workflow(graph=workflow.json) with three bound fn operators: compose_song (MiniMax-M3 LLM lyrics + structured caption), generate_song (@spaces.GPU xlarge worker, wav file dict + seed + stats), make_video (ffmpeg visualizer). All model machinery (AoTI kernels, StaticCache ladder, windowed DiT decode) unchanged. - workflow.json: references -> compose_song -> generate_song -> subjects, with audio fan-out into make_video -> Share Video. - app_blocks.py: original streaming Blocks app preserved. - README: hf_oauth: true for owner canvas editing.de1b9e6
