Music Creation Legacy

1import os2import sys3import json4import time5import random6import traceback78try:9    import requests10except ImportError:11    import subprocess12    subprocess.check_call([sys.executable, "-m", "pip", "install", "requests"])13    import requests1415COMFYUI_API_URL = os.getenv("COMFYUI_API_URL", "http://192.168.1.39:8188")16OUTPUT_DIR = "/data/output"1718# ---------- ACE-Step 1.5 music generation workflow ----------19WORKFLOW = {20    "3": {21        "inputs": {22            "seed": 31,23            "steps": 8,24            "cfg": 1,25            "sampler_name": "euler",26            "scheduler": "simple",27            "denoise": 1,28            "model": ["78", 0],29            "positive": ["94", 0],30            "negative": ["47", 0],31            "latent_image": ["98", 0],32        },33        "class_type": "KSampler",34        "_meta": {"title": "KSampler"},35    },36    "18": {37        "inputs": {38            "samples": ["3", 0],39            "vae": ["106", 0],40        },41        "class_type": "VAEDecodeAudio",42        "_meta": {"title": "VAE Decode Audio"},43    },44    "47": {45        "inputs": {46            "conditioning": ["94", 0],47        },48        "class_type": "ConditioningZeroOut",49        "_meta": {"title": "ConditioningZeroOut"},50    },51    "78": {52        "inputs": {53            "shift": 3,54            "model": ["104", 0],55        },56        "class_type": "ModelSamplingAuraFlow",57        "_meta": {"title": "ModelSamplingAuraFlow"},58    },59    "94": {60        "inputs": {61            "tags": ["109", 0],62            "lyrics": ["110", 0],63            "seed": 31,64            "bpm": 190,65            "duration": 90,66            "timesignature": "4",67            "language": "en",68            "keyscale": "E minor",69            "generate_audio_codes": True,70            "cfg_scale": 2,71            "temperature": 0.85,72            "top_p": 0.9,73            "top_k": 0,74            "min_p": 0.05,75            "clip": ["105", 0],76        },77        "class_type": "TextEncodeAceStepAudio1.5",78        "_meta": {"title": "TextEncodeAceStepAudio1.5"},79    },80    "98": {81        "inputs": {82            "seconds": 90,83            "batch_size": 1,84        },85        "class_type": "EmptyAceStep1.5LatentAudio",86        "_meta": {"title": "Empty Ace Step 1.5 Latent Audio"},87    },88    "104": {89        "inputs": {90            "unet_name": "acestep_v1.5_turbo.safetensors",91            "weight_dtype": "default",92        },93        "class_type": "UNETLoader",94        "_meta": {"title": "Load Diffusion Model"},95    },96    "105": {97        "inputs": {98            "clip_name1": "qwen_0.6b_ace15.safetensors",99            "clip_name2": "qwen_1.7b_ace15.safetensors",100            "type": "ace",101            "device": "default",102        },103        "class_type": "DualCLIPLoader",104        "_meta": {"title": "DualCLIPLoader"},105    },106    "106": {107        "inputs": {108            "vae_name": "ace_1.5_vae.safetensors",109        },110        "class_type": "VAELoader",111        "_meta": {"title": "Load VAE"},112    },113    "107": {114        "inputs": {115            "filename_prefix": "audio/ComfyUI",116            "quality": "V0",117            "audioUI": "",118            "audio": ["18", 0],119        },120        "class_type": "SaveAudioMP3",121        "_meta": {"title": "Save Audio (MP3)"},122    },123    "109": {124        "inputs": {125            "text": "",126        },127        "class_type": "Text Multiline",128        "_meta": {"title": "Text Multiline (Prompt)"},129    },130    "110": {131        "inputs": {132            "text": "",133        },134        "class_type": "Text Multiline",135        "_meta": {"title": "Text Multiline (Lyrics)"},136    },137}138139140def build_workflow(141    prompt: str,142    lyrics: str,143    duration: int,144    language: str,145    time_signature: str,146    bpm: int,147    keyscale: str,148) -> dict:149    """Build a music generation workflow with the given parameters."""150    import copy151152    wf = copy.deepcopy(WORKFLOW)153154    # Music description / tags (node 109)155    wf["109"]["inputs"]["text"] = prompt156157    # Lyrics (node 110)158    wf["110"]["inputs"]["text"] = lyrics159160    # ACE-Step encoder settings (node 94)161    wf["94"]["inputs"]["duration"] = duration162    wf["94"]["inputs"]["language"] = language163    wf["94"]["inputs"]["timesignature"] = time_signature164    wf["94"]["inputs"]["bpm"] = bpm165    wf["94"]["inputs"]["keyscale"] = keyscale166167    # Latent audio duration (node 98) — keep in sync with node 94168    wf["98"]["inputs"]["seconds"] = duration169170    # Randomize seeds (nodes 3 and 94)171    seed = random.randint(0, 2**31 - 1)172    wf["3"]["inputs"]["seed"] = seed173    wf["94"]["inputs"]["seed"] = seed174175    # Output prefix (node 107)176    wf["107"]["inputs"]["filename_prefix"] = "audio/emblema-music"177178    return wf179180181def submit_prompt(workflow: dict) -> str:182    """Submit workflow to ComfyUI and return prompt_id."""183    resp = requests.post(184        f"{COMFYUI_API_URL}/prompt",185        json={"prompt": workflow},186        timeout=30,187    )188    if resp.status_code != 200:189        try:190            error_detail = resp.json()191        except Exception:192            error_detail = resp.text193        raise RuntimeError(194            f"ComfyUI /prompt returned {resp.status_code}: "195            f"{json.dumps(error_detail, indent=2) if isinstance(error_detail, dict) else error_detail}"196        )197    data = resp.json()198199    # ComfyUI returns 200 even when nodes have validation errors200    node_errors = data.get("node_errors", {})201    if node_errors:202        raise RuntimeError(203            f"ComfyUI workflow has node errors: {json.dumps(node_errors, indent=2)}"204        )205206    return data["prompt_id"]207208209def wait_for_result(prompt_id: str, timeout: int = 900, poll_interval: int = 3) -> dict:210    """Poll ComfyUI history until the prompt completes with outputs."""211    deadline = time.time() + timeout212    empty_complete_retries = 0213    max_empty_retries = 3  # grace period for output serialization lag214215    while time.time() < deadline:216        resp = requests.get(217            f"{COMFYUI_API_URL}/history/{prompt_id}",218            timeout=10,219        )220        resp.raise_for_status()221        history = resp.json()222223        if prompt_id in history:224            prompt_data = history[prompt_id]225            status = prompt_data.get("status", {})226227            if status.get("status_str") == "error":228                messages = status.get("messages", [])229                raise RuntimeError(230                    f"ComfyUI prompt failed: {json.dumps(messages, indent=2)}"231                )232233            if status.get("completed", False):234                if prompt_data.get("outputs"):235                    return prompt_data236237                # Completed but no outputs — retry briefly for race condition238                empty_complete_retries += 1239                if empty_complete_retries >= max_empty_retries:240                    raise RuntimeError(241                        f"ComfyUI prompt completed but produced no outputs. "242                        f"This usually means a node failed silently (missing custom node or model). "243                        f"Status: {json.dumps(status, indent=2)}"244                    )245246        time.sleep(poll_interval)247248    raise TimeoutError(f"ComfyUI prompt {prompt_id} did not complete within {timeout}s")249250251def download_output_audio(prompt_data: dict, output_dir: str) -> str:252    """Download the generated audio from ComfyUI."""253    outputs = prompt_data.get("outputs", {})254    for node_id, node_output in outputs.items():255        audio_list = node_output.get("audio") or []256        if audio_list:257            audio_info = audio_list[0]258            filename = audio_info["filename"]259            subfolder = audio_info.get("subfolder", "")260            audio_type = audio_info.get("type", "output")261262            resp = requests.get(263                f"{COMFYUI_API_URL}/view",264                params={265                    "filename": filename,266                    "subfolder": subfolder,267                    "type": audio_type,268                },269                timeout=120,270            )271            resp.raise_for_status()272273            out_filename = f"generated_{filename}"274            out_path = os.path.join(output_dir, out_filename)275            with open(out_path, "wb") as f:276                f.write(resp.content)277278            return out_filename279280    raise RuntimeError(281        f"No output audio found in ComfyUI response. Available outputs: {json.dumps(outputs, indent=2)}"282    )283284285def main():286    try:287        input_json = sys.stdin.read()288        execution_input = json.loads(input_json)289        inputs = execution_input.get("inputs", {})290291        prompt = inputs.get("prompt", "")292        lyrics = inputs.get("lyrics", "")293        duration = int(inputs.get("duration", 90))294        language = inputs.get("language", "en")295        time_signature = inputs.get("time_signature", "4")296        bpm = int(inputs.get("bpm", 190))297        keyscale = inputs.get("keyscale", "E minor")298299        if not prompt:300            raise ValueError("Prompt input is required")301        if not lyrics:302            raise ValueError("Lyrics input is required")303        if not (10 <= duration <= 300):304            raise ValueError(f"Duration must be between 10 and 300, got {duration}")305        if not (40 <= bpm <= 300):306            raise ValueError(f"BPM must be between 40 and 300, got {bpm}")307308        os.makedirs(OUTPUT_DIR, exist_ok=True)309310        # Build workflow, submit, wait, download311        workflow = build_workflow(prompt, lyrics, duration, language, time_signature, bpm, keyscale)312        prompt_id = submit_prompt(workflow)313        prompt_data = wait_for_result(prompt_id)314        out_filename = download_output_audio(prompt_data, OUTPUT_DIR)315316        # Log metadata to stderr317        print(318            f"prompt_id={prompt_id}, duration={duration}s, bpm={bpm}, "319            f"key={keyscale}, time_sig={time_signature}, language={language}",320            file=sys.stderr,321        )322323        # Flat output — keys match OUTPUT_SCHEMA324        output = {325            "audio": out_filename,326        }327        print(json.dumps(output, indent=2))328329    except Exception as e:330        error_output = {331            "error": str(e),332            "errorType": type(e).__name__,333            "traceback": traceback.format_exc(),334        }335        print(json.dumps(error_output), file=sys.stderr)336        sys.exit(1)337338339if __name__ == "__main__":340    main()