{ "cells": [ { "cell_type": "code", "execution_count": null, "id": "fed7c348", "metadata": {}, "outputs": [], "source": [ "import os\n", "\n", "os.environ[\"CUDA_VISIBLE_DEVICES\"] = \"3\"" ] }, { "cell_type": "code", "execution_count": null, "id": "88c3355d", "metadata": {}, "outputs": [], "source": [ "from suno_utils.diffusion_stems import generation as diffusion_gen" ] }, { "cell_type": "code", "execution_count": null, "id": "c046f7b4", "metadata": {}, "outputs": [], "source": [ "import torch\n", "from suno_utils.audio import Audio\n", "\n", "_ = diffusion_gen.preload_models(compile=False)\n" ] }, { "cell_type": "markdown", "id": "fa37407c", "metadata": {}, "source": [ "Load a song to extract stems from" ] }, { "cell_type": "code", "execution_count": null, "id": "cf696991", "metadata": {}, "outputs": [], "source": [ "import json\n", "import numpy as np\n", "from suno_utils.utils.s3 import read_from_s3\n", "\n", "# gen_id = \"99bffa17-7e59-47b4-a048-5528cbda05d5\" # sister\n", "# gen_id = \"081d73c4-7805-4212-9c80-8db1137ca3c4\" # friends\n", "# gen_id = \"562f762d-6ced-4080-9af1-910ee3d0a5dc\" # something real\n", "# gen_id = \"23c15c62-494d-422d-8a60-8b0454044322\" # rubber duck\n", "# gen_id = \"4b140a9e-964b-422c-85b5-5861ad1a9d38\" # once\n", "# gen_id = \"7b214347-fa38-4e9b-96f4-f7ec65adea45\" # rock n roll\n", "gen_id = \"a5e2198a-f352-4abb-9a24-7f81b143ded3\" # stone\n", "# gen_id = \"7f5e7819-b7a1-449d-a76c-9c38950c4274\" # canada\n", "# gen_id = \"27135940-957f-4152-9e40-72e5c10a46bd\" # drunk\n", "# gen_id = \"caf1d7ae-ce81-4c75-b737-fd41e47a38ba\" # linger\n", "# gen_id = \"ac5aa29a-f18f-406f-a904-c47493f1beec\" # linger vox\n", "# gen_id = \"b04719c3-012a-4a4a-9b02-3cb2875d1744\" # courtney\n", "# gen_id = \"7407c7b3-8aa2-4f0a-82e8-fa19933f6e94\" # opera\n", "# gen_id = \"5b9624ef-b416-4095-a5f3-4945f09bba86\" # butterflies\n", "# gen_id = \"e757aa19-5c33-4945-aaa1-1519b7890f0b\" # bubblegum\n", "# gen_id = \"7f774078-1672-4858-a37f-acad373c5a84\" # golden\n", "# gen_id = \"1cee79db-a0ee-44e4-9646-d312cd620997\" # chinese\n", "\n", "s3_filepath = f\"s3://suno-data-uploads/studio/uploads/{gen_id}.npz\"\n", "mp3_filepath = f\"s3://suno-data-uploads/studio/uploads/{gen_id}.mp3\"\n", "vae_filepath = f\"s3://suno-data-uploads/studio/uploads/{gen_id}_vae.npz\"\n", "print(s3_filepath)\n", "data = read_from_s3(s3_filepath, read_f=np.load)\n", "vae_data = read_from_s3(vae_filepath, read_f=np.load)\n", "audio = Audio.from_s3(mp3_filepath, n_channels=2, sample_rate=48000)\n", "\n", "if \"v3.0_raw\" in data:\n", " codes = data[\"v3.0_raw\"]\n", "elif \"v3.5_raw\" in data:\n", " codes = data[\"v3.5_raw\"]\n", "elif \"v4.0_raw\" in data:\n", " codes = data[\"v4.0_raw\"]\n", "else:\n", " raise ValueError(\"No codes found\")\n", "\n", "text_data = read_from_s3(f\"s3://suno-data-uploads/studio/uploads/{gen_id}_hoot.json\")\n", "aligned_lyrics = json.loads(text_data)\n", "print(aligned_lyrics)\n", "audio.normalize_volume().play()" ] }, { "cell_type": "code", "execution_count": null, "id": "41632475", "metadata": {}, "outputs": [], "source": [ "# dict of stem types\n", "diffusion_gen.STEM_TYPE_ID_TO_NAME\n" ] }, { "cell_type": "markdown", "id": "29897167", "metadata": {}, "source": [ "Extract a stem" ] }, { "cell_type": "code", "execution_count": null, "id": "c0c34cd3", "metadata": {}, "outputs": [], "source": [ "steps = 12\n", "\n", "gen_cfg = diffusion_gen.DiffusionGenerationConfig(\n", " audio=audio,\n", " stem_type_id=0,\n", " # stem_type_cfg_scale=2,\n", " steps=steps,\n", " seed=0,\n", ")\n", "\n", "pred_audio, l = diffusion_gen.generate(gen_cfg)\n", "pred_audio.play()\n" ] }, { "cell_type": "code", "execution_count": null, "id": "0a1dd833", "metadata": {}, "outputs": [], "source": [ "gen_cfg = diffusion_gen.DiffusionGenerationConfig(\n", " audio=audio,\n", " stem_type_id=0,\n", " stem_type_cfg_scale=2,\n", " task=\"remove\",\n", " steps=steps,\n", " seed=0,\n", ")\n", "\n", "pred_audio, l = diffusion_gen.generate(gen_cfg)\n", "pred_audio.play()" ] }, { "cell_type": "code", "execution_count": null, "id": "6aa4e476", "metadata": {}, "outputs": [], "source": [ "gen_cfg = diffusion_gen.DiffusionGenerationConfig(\n", " audio=audio,\n", " stem_type_id=18,\n", " stem_type_cfg_scale=2,\n", " task=\"add\",\n", " steps=steps,\n", " seed=0,\n", ")\n", "\n", "pred_audio, l = diffusion_gen.generate(gen_cfg)\n", "pred_audio.play()" ] }, { "cell_type": "markdown", "id": "fe9f49bd", "metadata": {}, "source": [ "Dry a stem" ] }, { "cell_type": "code", "execution_count": null, "id": "86ebd4e6", "metadata": {}, "outputs": [], "source": [ "gen_id = \"f8a50e25-d705-49fa-b83a-82521014cb63\" # vocal\n", "lead_vocal = Audio.from_s3(gen_id)\n", "gen_cfg = diffusion_gen.DiffusionGenerationConfig(\n", " audio=lead_vocal,\n", " task=\"dry\",\n", " steps=steps,\n", " seed=0,\n", ")\n", "\n", "pred_audio, l = diffusion_gen.generate(gen_cfg)\n", "pred_audio.play()" ] }, { "cell_type": "markdown", "id": "4483a01d", "metadata": {}, "source": [ "Example modal call" ] }, { "cell_type": "code", "execution_count": null, "id": "4a22177e", "metadata": {}, "outputs": [], "source": [ "import modal\n", "import json\n", "from uuid import uuid4\n", "\n", "model_f_stem = modal.Cls.lookup(\"stems-stems_v1-dev\", \"StemStub\")\n", "gen_id = \"a5e2198a-f352-4abb-9a24-7f81b143ded3\" # stone\n", "\n", "input = {\n", " \"id\": str(uuid4()),\n", " \"prompt_audio\": gen_id,\n", " \"model_name\": \"stems\",\n", " \"callback_url\": \"https://api-staging.suno.ai/api/generate/finish-clip/\",\n", " \"metadata\": {\n", " \"stem_type_id\": 0,\n", " \"stem_type_cfg_scale\": 2,\n", " \"task\": \"stem\",\n", " \"stem_task\": \"remove\",\n", " },\n", "}\n", "\n", "print(f\"Starting modal call for {input['id']}\")\n", "model_f_stem.stem.remote(json.dumps(input))\n", "print(f\"Done with modal call for {input['id']}. See it at https://cdn1.suno.ai/{input['id']}.mp3\")\n" ] }, { "cell_type": "code", "execution_count": null, "id": "ff1f1925", "metadata": {}, "outputs": [], "source": [] } ], "metadata": { "language_info": { "name": "python" } }, "nbformat": 4, "nbformat_minor": 5 }