Tighten YuE for 16GB VRAM and stop the false Waiting status.

Cap Stage-A segments and use a heavier mmgp profile so long lyrics stop OOMing mid-run, and show that Comfy is actually working.

Co-authored-by: Cursor <cursoragent@cursor.com>

server/utils/musicWorkflow.ts

@@ -116,7 +116,9 @@ function buildYueWorkflow(params: MusicWorkflowParams): WorkflowGraph {
   const prefix = params.filenamePrefix || musicFilenamePrefix()
   const tags = params.tags
   const lyrics = params.lyrics || '[instrumental]\n'
-  const segments = Math.max(1, Math.min(10, lyrics.split(/\[[^\]]+\]/g).filter(part => part.trim()).length || 2))
+  // 5080 16GB OOMs past ~3 long Stage-A segments on fp16+mmgp. Cap hard.
+  const sections = lyrics.split(/\[[^\]]+\]/g).filter(part => part.trim()).length || 2
+  const segments = Math.max(1, Math.min(2, sections))
   for (const node of Object.values(graph)) {
     if (node.class_type === 'YUE_Stage_A_Loader') {
       node.inputs.stage_A_repo = yueStageARepo()
@@ -128,8 +130,9 @@ function buildYueWorkflow(params: MusicWorkflowParams): WorkflowGraph {
       node.inputs.genres_prompt = tags.replace(/,/g, ' ').replace(/\s+/g, ' ').trim()
       node.inputs.lyrics_prompt = lyrics
       node.inputs.seed = params.seed
-      node.inputs.prompt_end_time = params.duration
+      node.inputs.prompt_end_time = Math.min(params.duration, 45)
       node.inputs.run_n_segment = segments
+      node.inputs.max_new_tokens = 2000
     }
   }
   applyPrefix(graph, prefix)