Remove all audio decoding: generate becomes OUTPUT_NODE copying native MP3/WAV via shutil

Co-authored-by: lmangani <1423657+lmangani@users.noreply.github.com>

Author: Copilot

nodes.py

@@ -5,12 +5,8 @@
 import shutil
 import subprocess
 import tempfile
-import wave
 from typing import Any, Dict, List, Optional, Tuple, Union
 
-import numpy as np
-import torch
-
 import folder_paths
 
 CONFIG_FILE = os.path.join(os.path.dirname(__file__), "config.json")
@@ -1128,8 +1124,8 @@ def INPUT_TYPES(cls):
             },
         }
 
-    RETURN_TYPES = ("AUDIO",)
-    RETURN_NAMES = ("audio",)
+    RETURN_TYPES = ()
+    OUTPUT_NODE = True
     FUNCTION = "generate"
     CATEGORY = "AcestepCPP"
 
@@ -1394,33 +1390,14 @@ def generate(
                     f"stdout: {dit_result.stdout}\nstderr: {dit_result.stderr}"
                 )
 
-            # Decode the generated audio to raw PCM via ffmpeg, then read it
-            # with the built-in wave module.  This avoids optional torchaudio
-            # backends such as torchcodec that may not be installed.
-            decoded_wav = os.path.join(tmpdir, "decoded.wav")
-            _decode = subprocess.run(
-                ["ffmpeg", "-y", "-i", audio_path, "-acodec", "pcm_f32le", decoded_wav],
-                capture_output=True,
+            # Copy the native MP3/WAV to ComfyUI's output directory unchanged —
+            # no decoding, no re-encoding, no external tools.
+            out_dir = folder_paths.get_output_directory()
+            full_out_folder, base_name, counter, subfolder, _ = (
+                folder_paths.get_save_image_path("acestep", out_dir)
             )
-            if _decode.returncode != 0:
-                raise RuntimeError(
-                    f"ffmpeg could not decode {os.path.basename(audio_path)}:\n"
-                    + _decode.stderr.decode(errors="replace")
-                )
-
-            with wave.open(decoded_wav, "rb") as _wf:
-                n_ch = _wf.getnchannels()
-                sample_rate = _wf.getframerate()
-                raw = _wf.readframes(_wf.getnframes())
-
-            # pcm_f32le: 32-bit float, little-endian, interleaved channels
-            pcm = np.frombuffer(raw, dtype=np.float32)
-            if n_ch > 1:
-                pcm_2d = pcm.reshape(-1, n_ch).T.copy()  # (channels, samples)
-            else:
-                pcm_2d = pcm.reshape(1, -1).copy()       # (1, samples)
-            waveform = torch.from_numpy(pcm_2d)
-            # ComfyUI AUDIO format: waveform shape (batch, channels, samples)
-            audio = {"waveform": waveform.unsqueeze(0), "sample_rate": sample_rate}
+            save_filename = f"{base_name}_{counter:05}{ext}"
+            save_path = os.path.join(full_out_folder, save_filename)
+            shutil.copy2(audio_path, save_path)
 
-        return (audio,)
+        return {"ui": {"audio": [{"filename": save_filename, "subfolder": subfolder, "type": "output"}]}}

tests/test_nodes.py

@@ -333,7 +333,10 @@ def test_reference_audio_removed(self):
         assert "reference_audio_input" not in opt
 
     def test_return_types(self):
-        assert nodes.AcestepCPPGenerate.RETURN_TYPES == ("AUDIO",)
+        assert nodes.AcestepCPPGenerate.RETURN_TYPES == ()
+
+    def test_is_output_node(self):
+        assert nodes.AcestepCPPGenerate.OUTPUT_NODE is True
 
     def test_lego_tracks_list(self):
         """LEGO_TRACKS must include the track names documented in the README."""

tests/test_workflows.py

@@ -77,31 +77,24 @@ def test_has_generate_node(self, workflow_path):
         assert _nodes_by_type(wf, "AcestepCPPGenerate"), \
             "Workflow must include AcestepCPPGenerate"
 
-    # --- Output node: save-then-play --------------------------------------
+    # --- Output: generate node is itself the output node ---------------------
 
-    def test_uses_save_audio_not_preview_audio(self, workflow_path):
+    def test_has_no_audio_output_node(self, workflow_path):
+        """AcestepCPPGenerate is itself an OUTPUT_NODE; no SaveAudio or PreviewAudio needed."""
         wf = _load(workflow_path)
         node_types = [n["type"] for n in wf["nodes"]]
+        assert "SaveAudio" not in node_types, \
+            "SaveAudio is no longer needed; AcestepCPPGenerate copies the file and shows its own player"
         assert "PreviewAudio" not in node_types, \
-            "PreviewAudio (play-only) should be replaced by SaveAudio (save-then-play)"
-        assert "SaveAudio" in node_types, \
-            "SaveAudio must be present so generated audio is persisted to disk"
+            "PreviewAudio is no longer needed; AcestepCPPGenerate shows its own audio player"
 
-    def test_save_audio_has_filename_prefix(self, workflow_path):
+    def test_generate_has_no_output_links(self, workflow_path):
+        """AcestepCPPGenerate is a terminal output node; its output slot must have no outgoing links."""
         wf = _load(workflow_path)
-        for node in _nodes_by_type(wf, "SaveAudio"):
-            wv = node.get("widgets_values", [])
-            assert len(wv) >= 1 and isinstance(wv[0], str) and wv[0], \
-                "SaveAudio must have a non-empty filename_prefix widget value"
-
-    def test_save_audio_format_is_mp3(self, workflow_path):
-        """SaveAudio must use mp3 format — not the ComfyUI default of flac."""
-        wf = _load(workflow_path)
-        for node in _nodes_by_type(wf, "SaveAudio"):
-            wv = node.get("widgets_values", [])
-            got = repr(wv[1]) if len(wv) > 1 else "(missing)"
-            assert len(wv) >= 2 and wv[1] == "mp3", \
-                f"SaveAudio format (index 1) must be 'mp3', got {got}"
+        for node in _nodes_by_type(wf, "AcestepCPPGenerate"):
+            for output in node.get("outputs", []):
+                assert not output.get("links"), \
+                    f"AcestepCPPGenerate output '{output.get('name')}' should have no outgoing links"
 
     # --- Model loader widget values ---------------------------------------
 
@@ -310,12 +303,3 @@ def test_model_loader_output_connected(self, workflow_path):
         for node in _nodes_by_type(wf, "AcestepCPPModelLoader"):
             assert node["id"] in link_src_nodes, \
                 "AcestepCPPModelLoader output is not connected to any node"
-
-    def test_generate_output_connected_to_save_audio(self, workflow_path):
-        """AcestepCPPGenerate audio output must feed into SaveAudio."""
-        wf = _load(workflow_path)
-        node_ids = {n["id"]: n for n in wf["nodes"]}
-        save_audio_ids = {n["id"] for n in _nodes_by_type(wf, "SaveAudio")}
-        # Walk links: find any link whose destination is a SaveAudio node
-        connected = any(lnk[3] in save_audio_ids for lnk in wf.get("links", []))
-        assert connected, "No link from any node to SaveAudio found"

workflow-examples/acestep-cpp-cover.json

@@ -127,16 +127,7 @@
           "link": null
         }
       ],
-      "outputs": [
-        {
-          "name": "audio",
-          "type": "AUDIO",
-          "links": [
-            2
-          ],
-          "slot_index": 0
-        }
-      ],
+      "outputs": [],
       "properties": {
         "Node name for S&R": "AcestepCPPGenerate"
       },
@@ -167,36 +158,6 @@
         "",
         1.0
       ]
-    },
-    {
-      "id": 3,
-      "type": "SaveAudio",
-      "pos": [
-        880,
-        20
-      ],
-      "size": {
-        "0": 320,
-        "1": 130
-      },
-      "flags": {},
-      "order": 3,
-      "mode": 0,
-      "inputs": [
-        {
-          "name": "audio",
-          "type": "AUDIO",
-          "link": 2
-        }
-      ],
-      "outputs": [],
-      "properties": {
-        "Node name for S&R": "SaveAudio"
-      },
-      "widgets_values": [
-        "acestep/output",
-        "mp3"
-      ]
     }
   ],
   "links": [
@@ -208,14 +169,6 @@
       0,
       "ACESTEP_MODELS"
     ],
-    [
-      2,
-      2,
-      0,
-      3,
-      0,
-      "AUDIO"
-    ],
     [
       3,
       4,

workflow-examples/acestep-cpp-lora.json

@@ -128,16 +128,7 @@
           "link": null
         }
       ],
-      "outputs": [
-        {
-          "name": "audio",
-          "type": "AUDIO",
-          "links": [
-            2
-          ],
-          "slot_index": 0
-        }
-      ],
+      "outputs": [],
       "properties": {
         "Node name for S&R": "AcestepCPPGenerate"
       },
@@ -168,36 +159,6 @@
         "",
         1.0
       ]
-    },
-    {
-      "id": 3,
-      "type": "SaveAudio",
-      "pos": [
-        880,
-        20
-      ],
-      "size": {
-        "0": 320,
-        "1": 130
-      },
-      "flags": {},
-      "order": 3,
-      "mode": 0,
-      "inputs": [
-        {
-          "name": "audio",
-          "type": "AUDIO",
-          "link": 2
-        }
-      ],
-      "outputs": [],
-      "properties": {
-        "Node name for S&R": "SaveAudio"
-      },
-      "widgets_values": [
-        "acestep/output",
-        "mp3"
-      ]
     }
   ],
   "links": [
@@ -209,14 +170,6 @@
       0,
       "ACESTEP_MODELS"
     ],
-    [
-      2,
-      2,
-      0,
-      3,
-      0,
-      "AUDIO"
-    ],
     [
       3,
       4,

workflow-examples/acestep-cpp-reference-audio.json

@@ -127,16 +127,7 @@
           "link": null
         }
       ],
-      "outputs": [
-        {
-          "name": "audio",
-          "type": "AUDIO",
-          "links": [
-            2
-          ],
-          "slot_index": 0
-        }
-      ],
+      "outputs": [],
       "properties": {
         "Node name for S&R": "AcestepCPPGenerate"
       },
@@ -167,36 +158,6 @@
         "",
         1.0
       ]
-    },
-    {
-      "id": 3,
-      "type": "SaveAudio",
-      "pos": [
-        880,
-        20
-      ],
-      "size": {
-        "0": 320,
-        "1": 130
-      },
-      "flags": {},
-      "order": 3,
-      "mode": 0,
-      "inputs": [
-        {
-          "name": "audio",
-          "type": "AUDIO",
-          "link": 2
-        }
-      ],
-      "outputs": [],
-      "properties": {
-        "Node name for S&R": "SaveAudio"
-      },
-      "widgets_values": [
-        "acestep/output",
-        "mp3"
-      ]
     }
   ],
   "links": [
@@ -208,14 +169,6 @@
       0,
       "ACESTEP_MODELS"
     ],
-    [
-      2,
-      2,
-      0,
-      3,
-      0,
-      "AUDIO"
-    ],
     [
       3,
       4,