aigen/server/utils/musicWorkflow.ts

import workflowAceStep from '../assets/workflow_ace_step.json'
import workflowAceStep15 from '../assets/workflow_ace_step_15.json'
import workflowYue from '../assets/workflow_yue.json'
import type { MusicEngine } from '~/utils/music'
import { parseMusicBpm } from '~/utils/music'

export type MusicWorkflowParams = {
  tags: string
  lyrics: string
  duration: number
  steps: number
  seed: number
  cfg: number
  lyricsStrength: number
  samplerName?: string
  scheduler?: string
  filenamePrefix?: string
  checkpoint?: string
  engine?: MusicEngine
  vocalMultiplier?: number
  bpm?: number
}

type WorkflowNode = { class_type: string; inputs: Record<string, unknown>; _meta?: { title?: string } }
type WorkflowGraph = Record<string, WorkflowNode>

const ACE_V1_CKPT = 'ace_step_v1_3.5b.safetensors'
const ACE_15_CKPT = 'ace_step_1.5_turbo_aio.safetensors'
const YUE_S1_DEFAULT = 'C:\\Users\\ianjm\\AppData\\Local\\Comfy-Desktop\\ComfyUI-Shared\\models\\yue\\YuE-s1-7B-anneal-en-cot'
const YUE_S2_DEFAULT = 'C:\\Users\\ianjm\\AppData\\Local\\Comfy-Desktop\\ComfyUI-Shared\\models\\yue\\YuE-s2-1B-general'
const YUE_NODE = 'YUE_Stage_A_Loader'

export function musicFilenamePrefix() {
  return String(useRuntimeConfig().comfyAudioPrefix || process.env.COMFY_AUDIO_PREFIX || 'audio/AIGen')
}

function yueStageARepo() {
  return String(process.env.YUE_STAGE_A || useRuntimeConfig().yueStageA || YUE_S1_DEFAULT)
}

function yueStageBRepo() {
  return String(process.env.YUE_STAGE_B || useRuntimeConfig().yueStageB || YUE_S2_DEFAULT)
}

function applyPrefix(graph: WorkflowGraph, prefix: string) {
  for (const node of Object.values(graph)) {
    if ((node.class_type === 'SaveAudio' || node.class_type === 'SaveAudioMP3') && 'filename_prefix' in node.inputs) {
      node.inputs.filename_prefix = prefix
    }
  }
}

function buildAceV1Workflow(params: MusicWorkflowParams): WorkflowGraph {
  const graph = structuredClone(workflowAceStep as WorkflowGraph)
  const prefix = params.filenamePrefix || musicFilenamePrefix()
  const sampler = params.samplerName || 'euler'
  const scheduler = params.scheduler || 'simple'
  for (const node of Object.values(graph)) {
    if (node.class_type === 'CheckpointLoaderSimple' && 'ckpt_name' in node.inputs) {
      node.inputs.ckpt_name = params.checkpoint || ACE_V1_CKPT
    }
    if (node.class_type === 'TextEncodeAceStepAudio') {
      node.inputs.tags = params.tags
      node.inputs.lyrics = params.lyrics
      node.inputs.lyrics_strength = params.lyricsStrength
    }
    if (node.class_type === 'EmptyAceStepLatentAudio' && 'seconds' in node.inputs) {
      node.inputs.seconds = params.duration
    }
    if (node.class_type === 'KSampler') {
      node.inputs.seed = params.seed
      node.inputs.steps = params.steps
      node.inputs.cfg = params.cfg
      node.inputs.sampler_name = sampler
      node.inputs.scheduler = scheduler
    }
  }
  applyPrefix(graph, prefix)
  return graph
}

function buildAce15Workflow(params: MusicWorkflowParams): WorkflowGraph {
  const graph = structuredClone(workflowAceStep15 as WorkflowGraph)
  const prefix = params.filenamePrefix || musicFilenamePrefix()
  const sampler = params.samplerName || 'euler'
  const scheduler = params.scheduler || 'simple'
  const bpm = params.bpm || parseMusicBpm(params.tags)
  for (const node of Object.values(graph)) {
    if (node.class_type === 'CheckpointLoaderSimple' && 'ckpt_name' in node.inputs) {
      node.inputs.ckpt_name = params.checkpoint || ACE_15_CKPT
    }
    if (node.class_type === 'TextEncodeAceStepAudio1.5') {
      node.inputs.tags = params.tags
      node.inputs.lyrics = params.lyrics
      node.inputs.seed = params.seed
      node.inputs.bpm = bpm
      node.inputs.duration = params.duration
    }
    if (node.class_type === 'EmptyAceStep1.5LatentAudio' && 'seconds' in node.inputs) {
      node.inputs.seconds = params.duration
    }
    if (node.class_type === 'KSampler') {
      node.inputs.seed = params.seed
      node.inputs.steps = params.steps
      node.inputs.cfg = params.cfg
      node.inputs.sampler_name = sampler
      node.inputs.scheduler = scheduler
    }
  }
  applyPrefix(graph, prefix)
  return graph
}

function buildYueWorkflow(params: MusicWorkflowParams): WorkflowGraph {
  const graph = structuredClone(workflowYue as WorkflowGraph)
  const prefix = params.filenamePrefix || musicFilenamePrefix()
  const tags = params.tags
  const lyrics = params.lyrics || '[Instrumental]\n'
  // Beast 5080 16GB recipe (keep in sync with workflow_yue.json):
  // - mmgp profile 2 only (3–5 enable quanto and crash under Comfy InferenceMode)
  // - exactly 1 Stage-A lyric section (2+ still OOMs beside xcodec)
  // - max_new_tokens floor is 2944 (node min); do not go lower
  // - Triton/torch.compile off on Windows
  // - xcodec stays on CPU during Stage A (node patch)
  const segments = 1
  for (const node of Object.values(graph)) {
    if (node.class_type === 'YUE_Stage_A_Loader') {
      node.inputs.stage_A_repo = yueStageARepo()
      node.inputs.quantization_model = 'fp16'
      node.inputs.use_mmgp = true
      node.inputs.stage1_cache_size = 8192
      node.inputs.mmgp_profile = 2
    }
    if (node.class_type === 'YUE_Stage_B_Loader') {
      node.inputs.stage_B_repo = yueStageBRepo()
      node.inputs.use_mmgp = true
      node.inputs.stage2_cache_size = 4096
      node.inputs.stage2_batch_size = 1
    }
    if (node.class_type === 'YUE_Stage_A_Sampler') {
      // Genres go to the model prompt only. Stage-A npy filenames are short (Windows MAX_PATH).
      node.inputs.genres_prompt = tags.replace(/,/g, ' ').replace(/\s+/g, ' ').trim().slice(0, 400)
      node.inputs.lyrics_prompt = lyrics
      node.inputs.seed = params.seed
      node.inputs.prompt_end_time = Math.min(params.duration, 30)
      node.inputs.run_n_segment = segments
      node.inputs.max_new_tokens = 2944
      node.inputs.use_dual_tracks_prompt = false
      node.inputs.use_audio_prompt = false
      node.inputs.offload_model = true
      node.inputs.stage1_no_guidance = true
    }
  }
  applyPrefix(graph, prefix)
  return graph
}

export function buildMusicWorkflow(params: MusicWorkflowParams): WorkflowGraph {
  const engine = params.engine || 'ace-step'
  if (engine === 'yue') return buildYueWorkflow(params)
  if (engine === 'ace-step-1.5') return buildAce15Workflow(params)
  return buildAceV1Workflow(params)
}

export async function assertMusicEngineNodes(engine: MusicEngine | undefined) {
  const { comfyHasClassType } = await import('~/server/utils/comfy')
  if (engine === 'yue') {
    const present = await comfyHasClassType(YUE_NODE)
    if (present === false) {
      throw createError({
        statusCode: 400,
        statusMessage: 'YuE is not installed on Beast. Clone ComfyUI_YuE into custom_nodes and restart Comfy, then try again.'
      })
    }
  }
  if (engine === 'ace-step-1.5') {
    const present = await comfyHasClassType('TextEncodeAceStepAudio1.5')
    if (present === false) {
      throw createError({
        statusCode: 400,
        statusMessage: 'This ComfyUI build does not have ACE-Step 1.5 nodes. Update Comfy Desktop, or turn the 1.5 toggle off.'
      })
    }
  }
}

export const MUSIC_NODE_LABELS: Record<string, string> = Object.fromEntries([
  ...Object.entries(workflowAceStep as WorkflowGraph).map(([id, node]) => [id, node._meta?.title || `Node ${id}`]),
  ...Object.entries(workflowAceStep15 as WorkflowGraph).map(([id, node]) => [id, node._meta?.title || `Node ${id}`]),
  ...Object.entries(workflowYue as WorkflowGraph).map(([id, node]) => [id, node._meta?.title || `Node ${id}`])
])