Practical SOP: Generating Commercial-Ready Music with Suno and ACE-Step

AI music generation has two routes. One is the SaaS fast route: Suno or ElevenLabs Music, where you open a web page, type a prompt, and get a song in seconds, with paid plans granting commercial rights. The other is the local free route: ACE-Step, an MIT-licensed open-source model you run on your own machine, no subscription, no per-generation fee. Picking the wrong route wastes effort -- using Suno's free tier for commercial work (watermarked, non-commercial) or paying for Suno when ACE-Step runs free locally.

This SOP covers five steps: requirements and lyrics, choose tool, generate, post-production, integrate. It complements the batch's ACE-Step Open Source Project Analysis (deep dive into the local model) and AI Music Generation Tools Comparison (five-tool comparison): those answer "what and why," this one answers "how to do it." ACE-Step commands are verified against the official GitHub README; Suno and ElevenLabs per their official sites. It draws a clear line from the site's AI Podcast Production SOP -- this article is about music generation (songwriting), while the podcast SOP covers TTS voice.

1. Two Routes: Fast SaaS vs Free Local

The core decision is made before you start: do you want fast and convenient, or free and controllable?

In short: fast turnaround or paid commercial use goes to Suno/ElevenLabs; free or commercial-ready and willing to tinker goes to ACE-Step local.

2. The Five-Step SOP

Step 1: Requirements and Lyrics

Define four things before you start:

Style. Pop, electronic, folk, hip-hop, R&B? Style determines the tags you feed Suno and the prompt for ACE-Step. Be specific -- "90s Hong Kong pop" beats "pop" for hit rate.

BPM and duration. Beats per minute and length define rhythm and structure. Suno lets you set duration (typically 1-4 minutes); ACE-Step generates clips up to 4 minutes. For video scoring, align BPM with visual pacing.

Vocals or instrumental. A full song with vocals (lyrics plus melody plus singing), or instrumental BGM? Suno and ACE-Step both generate vocals; ElevenLabs Music supports vocals too. For instrumental-only, see Stable Audio in the site's AI Music Generation Tools Comparison.

Lyrics and copyright. Copyright ownership of AI-generated lyrics is a legal gray area. Most platform terms grant generated content to the user, but training-data copyright lawsuits (Suno and Udio sued by RIAA) are unresolved. Before commercial use: do not use copyrighted lyrics as prompts, do not assume "AI-generated means no copyright risk," and read the platform's latest terms.

Step 2: Choose Tool

Decide across three dimensions:

Commercial rights. Commercial use requires Suno or ElevenLabs paid plans (granting commercial rights) or ACE-Step (MIT, commercial OK). Suno's free tier gives 50 credits/day with watermark and non-commercial only -- do not use it for commercial content.

Cost. Zero budget plus a GPU means ACE-Step local, software free, only electricity. Zero budget, no GPU means Suno free tier for personal non-commercial testing. Budget available means Suno $8-10/mo or ElevenLabs Music subscription for better convenience and quality.

Quality and controllability. Max convenience and commercially polished quality goes to Suno. Model in hand, modifiable, secondary development goes to ACE-Step. For detailed selection, see the batch's AI Music Generation Tools Comparison.

Step 3: Generate

SaaS route (Suno): Open suno.com, enter a prompt (style description plus lyrics), click generate. Suno v4.5 supports custom lyrics, style tags, and vocal/instrumental toggle. Example prompt:

Style: 90s Hong Kong pop, male vocal, melancholic
Lyrics:
[Verse 1]
The city lights dim
[Chorus]
...

Generation takes seconds to a minute, producing a full song with vocals. Free tier: 50 credits/day (watermarked, non-commercial); Pro from $10/mo.

Local route (ACE-Step): Commands verified against the official GitHub README. First, set up the environment:

# 1. Clone the repo
git clone https://github.com/ace-step/ACE-Step.git
cd ACE-Step

# 2. Create a Python 3.10 environment (conda or venv)
conda create -n ace_step python=3.10 -y
conda activate ace_step
# Or with venv: python -m venv venv && source venv/bin/activate  # Windows: venv\Scripts\activate.bat

# 3. Install PyTorch (CUDA 12.6 example; Mac uses MPS and skips this)
pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu126

# 4. Install ACE-Step
pip install -e .

Launch the Gradio web UI:

acestep --port 7865

Or launch with options (checkpoint path, device, bf16 precision):

acestep --checkpoint_path /path/to/checkpoint --port 7865 --device_id 0 --share true --bf16 true

Open a browser at http://localhost:7865, enter a prompt (style plus lyrics), and click generate. On an A100, ACE-Step produces 4 minutes of music in about 20 seconds; an RTX 4090 renders 1 minute of audio in about 1.74 seconds (27 steps); a MacBook M2 Max takes about 26 seconds for 1 minute. Performance per the official README; actual speed depends on your hardware.

Step 4: Post-Production (Mixing, Tuning, Vocal Processing)

AI-generated music usually needs post-production before formal release. Three tasks:

Mixing. AI-generated vocals and accompaniment are often merged with uneven dynamics. Use Audacity (free, open source) or Reaper for EQ, compression, and limiting to balance overall volume.

Tuning. AI vocals may have pitch and timing offsets. Use open-source tools like AudioShake or paid tools like iZotope RX for vocal separation and tuning. ACE-Step supports lyric2vocal (lyrics to vocal) and singing2accompaniment (vocal to accompaniment), letting you export vocal and backing tracks separately for independent processing.

Vocal processing. Add reverb, delay, and harmony to give vocals space. Audacity includes basic effects; for professional-grade processing, use Reaper or Logic Pro.

Step 5: Integrate (Video, Podcast, Content)

Export by target scenario:

Video scoring. Short videos use MP3/AAC at 192kbps; long-form uses lossless WAV. Align BPM with visual rhythm (hit points in the editing software). Normalize volume (-14 LUFS for streaming, -16 LUFS for short video).

Podcast. For BGM or intro/outro music, export MP3 at 128-192kbps. When mixing with podcast vocals, keep BGM below -20dB to avoid overpowering speech. The site's AI Podcast Production SOP covers the TTS voice line; this article is about music generation -- the two complement each other.

Content distribution. Uploading to Spotify or Apple Music requires a distributor like DistroKid, format WAV 44.1kHz/16bit. Confirm a clear copyright chain before commercial release (paid plan commercial rights or MIT license).

3. Five Pitfalls

Pitfall 1: Free tier is non-commercial; commercial use invites liability. Suno's free tier gives 50 credits/day with watermark and explicitly forbids commercial use; ElevenLabs free tier has similar limits. Using free-tier music in commercial videos, streaming releases, or ads violates terms. Fix: buy a paid plan (Suno from $8-10/mo) or use ACE-Step locally (MIT, commercial OK). Do not push your luck.

Pitfall 2: Lyrics copyright is a gray area; AI-generated does not mean risk-free. Using copyrighted lyrics as a prompt for AI to re-sing may infringe the original lyric copyright. While most platform terms grant generated content to the user, training-data lawsuits (Suno and Udio sued by RIAA) remain unresolved. Fix: do not use copyrighted lyrics as prompts; read the platform's latest terms before commercial use; consult legal counsel if needed.

Pitfall 3: Quality is unstable; same prompt yields different results. AI music generation is random -- the same prompt can produce a stunning track once and a terrible one the next. Both Suno and ACE-Step have this issue. Fix: run multiple times and pick the best; use Suno's free credits to test parameters before a paid run; in ACE-Step, adjust sampling steps (27 for speed, 60 for quality) and random seeds.

Pitfall 4: Vocals and accompaniment are merged; cannot adjust separately. Suno's free tier outputs vocals and accompaniment in a single audio file, with no way to adjust vocals or swap backing tracks independently. Fix: use Audacity's vocal isolation (basic quality), or professional tools like iZotope RX or AudioShake for stem separation. ACE-Step supports lyric2vocal and singing2accompaniment, generating vocal and backing tracks separately for greater post-production flexibility.

Pitfall 5: Export format mismatches the target platform. Short video platforms want MP3/AAC; streaming services want WAV 44.1kHz/16bit; game engines want OGG. AI tools' default exports may not match. Fix: confirm the target platform's required format and bitrate before export; use Audacity or ffmpeg for format conversion. Normalize volume to platform standards (streaming -14 LUFS, short video -16 LUFS).

FAQ

Q1: Can I use Suno's free tier commercially? A1: No. Suno's free tier provides 50 credits per day, and generated music is watermarked and explicitly non-commercial. Commercial use requires a paid plan (Pro from $10/mo grants commercial rights) or the ACE-Step local route (MIT license, commercial OK). Do not use the free tier for commercial content.

Q2: What hardware does ACE-Step need to run locally? A2: A GPU is required. The official README supports Mac (Apple Silicon MPS), AMD, Intel, and CUDA devices. CPU is not usable. VRAM or unified memory should typically start at 8GB. Performance reference: RTX 4090 renders 1 minute of audio in about 1.74 seconds; MacBook M2 Max takes about 26 seconds. Without a GPU, Suno is the practical choice.

Q3: Who owns the copyright to AI-generated lyrics? A3: It is a legal gray area. Most platform terms grant generated content to the user, but training-data copyright lawsuits (Suno and Udio sued by RIAA) are unresolved. Do not use copyrighted lyrics as prompts; read the platform's latest terms before commercial use. ACE-Step's MIT license covers code and model, but the copyright of generated content still requires your own judgment.

Q4: How do I choose between Suno and ACE-Step? A4: It depends on commercial use, cost, and controllability. For fast, convenient, paid commercial use, choose Suno. For free, commercial-ready, willing to tinker, choose ACE-Step local. A common combo: Suno for quick demos and inspiration, ACE-Step for batch generation and deep customization. For a detailed comparison, see the batch's AI Music Generation Tools Comparison.

Q5: How is this different from the AI Podcast Production SOP? A5: This article is about music generation (songwriting), using Suno or ACE-Step to produce full songs with vocals or instrumental tracks. The Podcast SOP covers TTS voice (text-to-speech), using tools like ElevenLabs to read text aloud. The two complement each other: music generation for BGM and songs, TTS for podcast vocals.

References

• Suno official site (pricing, commercial terms, v4.5 features): https://suno.com
• ElevenLabs Music official site (pricing, genre switching): https://elevenlabs.ai
• ACE-Step GitHub repository (install commands, Gradio UI, benchmarks): https://github.com/ace-step/ACE-Step
• ACE-Step official site (acemusic.ai): https://acemusic.ai
• This site's AI Music Generation Tools Comparison: https://aiwebcool.com/en/ai-music-generation-tools-comparison-review