> ## Documentation Index > Fetch the complete documentation index at: https://docs.modelslab.com/llms.txt > Use this file to discover all available pages before exploring further. # Song Generation Guide > Professional guide to creating music with ModelsLab Song Generator API powered by ACE-Step v1.5 Song Generation Guide # Song Generation Guide This guide contains professional music creation knowledge to help you create high-quality songs using the ModelsLab Song Generator API, powered by **ACE-Step v1.5** model. *** ## Overview The Song Generator API allows you to create complete songs with vocals in 50+ languages using the advanced **ACE-Step v1.5** model. You can either provide your own lyrics or let the AI generate them automatically based on your prompt. ### Key Features * **Duration Control**: Generate songs from 30 seconds to 8 minutes (30-480 seconds) * **50+ Languages**: Support for languages from Arabic to Chinese * **Lyrics Generation**: Automatic lyrics generation or use your own * **Instrumental Mode**: Generate instrumental versions without vocals * **Style Control**: Use caption to define music style, instruments, and atmosphere *** ## Understanding the Parameters ### Caption: Your Music Blueprint **Caption is the most important parameter** affecting your generated song. It describes the overall music elements you want. #### What to Include in Caption | Dimension | Examples | | - | - | | **Style/Genre** | pop, rock, jazz, electronic, hip-hop, R\&B, folk, reggaeton, synthwave | | **Emotion/Atmosphere** | melancholic, uplifting, energetic, dreamy, dark, nostalgic, euphoric, intimate | | **Instruments** | acoustic guitar, piano, synth pads, 808 drums, strings, brass, electric bass | | **Timbre Texture** | warm, bright, crisp, airy, punchy, lush, raw, polished | | **Era Reference** | 80s synth-pop, 90s grunge, 2010s EDM, vintage soul, modern trap | | **Vocal Characteristics** | female vocal, male vocal, breathy, powerful, falsetto, raspy | | **Production Style** | lo-fi, high-fidelity, live recording, studio-polished | #### Caption Writing Principles 1. **Be Specific** — "sad piano ballad with female breathy vocal" works better than "a sad song" 2. **Combine Dimensions** — Mix style + emotion + instruments + timbre for precise control 3. **Use References** — "80s synthwave style" or "reggaeton with flamenco influence" conveys complex aesthetics quickly 4. **Texture Words Matter** — Adjectives like warm, crisp, airy, punchy influence mixing and timbre 5. **Balance Detail vs Freedom** — More details = more control, fewer details = more AI creativity 6. **Avoid Conflicts** — Don't combine incompatible styles like "classical strings" and "hardcore metal" unless you want evolution **Example Good Captions:** ``` A modern reggaeton track with strong flamenco influence, featuring female vocal with reverb, deep sub-bass, crisp percussion, and plucked synth guitar riff ``` ``` Lo-fi hip-hop beat with warm vinyl crackle, mellow piano chords, subtle jazz drums, and atmospheric pad textures ``` ``` 80s synthwave pop with bright synth leads, punchy drum machine, nostalgic atmosphere, and powerful female vocals ``` *** ## Lyrics: Your Song's Timeline Lyrics control how your song unfolds over time. They include: * Lyric text content * Structure tags (\[Verse], \[Chorus], etc.) * Vocal style hints * Instrumental sections * Energy changes ### Common Structure Tags | Category | Tag | Description | | - | - | - | | **Basic Structure** | `[Intro]` | Opening, establish atmosphere | | | `[Verse]` / `[Verse 1]` | Verse, narrative progression | | | `[Pre-Chorus]` | Build energy before chorus | | | `[Chorus]` | Emotional climax, hook | | | `[Bridge]` | Transition or elevation | | | `[Outro]` | Ending, conclusion | | **Dynamic Sections** | `[Build]` | Energy gradually rising | | | `[Drop]` | Electronic music energy release | | | `[Breakdown]` | Reduced instrumentation | | **Instrumental** | `[Instrumental]` | Pure instrumental, no vocals | | | `[Guitar Solo]` | Guitar solo section | | | `[Piano Interlude]` | Piano interlude | | **Special Tags** | `[Fade Out]` | Fade out ending | ### Vocal Control Tags | Tag | Effect | | - | - | | `[raspy vocal]` | Raspy, textured vocals | | `[whispered]` | Whispered vocals | | `[falsetto]` | Falsetto vocals | | `[powerful belting]` | Powerful, high-pitched singing | | `[harmonies]` | Layered harmonies | ### Energy Tags | Tag | Effect | | - | - | | `[high energy]` | High energy, passionate | | `[building energy]` | Increasing energy | | `[explosive]` | Explosive energy | | `[melancholic]` | Melancholic mood | | `[euphoric]` | Euphoric feeling | ### Lyrics Writing Tips **1. Control Syllable Count** Keep **6-10 syllables per line** for best results. Consistent syllable counts create better rhythm. **2. Use Case for Intensity** ``` [Verse] walking through the empty streets (normal) [Chorus] WE ARE THE CHAMPIONS! (high intensity) ``` **3. Parentheses for Background Vocals** ``` [Chorus] We rise together (together) Into the light (into the light) ``` **4. Clear Section Separation** Always separate sections with blank lines: ``` [Verse 1] First verse lyrics here Continue first verse [Chorus] Chorus lyrics here Chorus continues ``` ### Keep Caption and Lyrics Consistent ⚠️ **Critical**: Descriptions in Caption and Lyrics must align. If Caption says "soft piano ballad" but Lyrics has `[explosive metal solo]`, results will be poor. **Checklist:** * Instruments in Caption ↔ Instrumental tags in Lyrics * Emotion in Caption ↔ Energy tags in Lyrics * Vocal description in Caption ↔ Vocal control tags in Lyrics *** ## Duration Calculation **You MUST calculate appropriate duration** based on your lyrics and structure. ### Estimation Method * **Per line of lyrics**: 3-5 seconds * **Intro/Outro**: 5-10 seconds each * **Instrumental sections**: 5-15 seconds * **Typical structures**: * 2 verses + 2 choruses: 120-150 seconds minimum * 2 verses + 2 choruses + bridge: 180-240 seconds * Full song with intro/outro: 210-270 seconds (3.5-4.5 minutes) ### Common Pitfall ❌ **DON'T**: 10 lines of lyrics with 60 seconds duration → rushed and compressed ✅ **DO**: 10 lines → \~40 seconds vocals + 20 seconds intro/outro = 60+ seconds **Rule**: When in doubt, estimate longer rather than shorter. *** ## Using Lyrics Generation When you don't have lyrics, set `lyrics_generation: true` and provide: 1. **prompt**: Describe the topic/theme for lyrics 2. **caption**: Describe the music style (same as with manual lyrics) ### Example Request with Lyrics Generation ```json theme={"theme":{"light":"github-light","dark":"github-dark"}} { "key": "your_api_key", "lyrics_generation": true, "prompt": "A song about overcoming challenges and finding inner strength, with uplifting message and emotional journey from doubt to confidence", "caption": "Inspiring pop ballad with piano and strings, building from intimate verse to powerful anthemic chorus, female vocal with emotional delivery", "duration": 180, "webhook": null, "track_id": null } ``` *** ## Instrumental Mode To generate music without vocals, set `instrumental: true`: ```json theme={"theme":{"light":"github-light","dark":"github-dark"}} { "key": "your_api_key", "lyrics_generation": false, "lyrics": "[Instrumental]", "caption": "Energetic electronic dance music with driving bassline, synth melodies, and dynamic build-ups", "instrumental": true, "duration": 240, "webhook": null, "track_id": null } ``` *** ## Language Support The API supports 50+ languages. Specify the language code: | Language | Code | Language | Code | | - | - | - | - | | English | en | Spanish | es | | Chinese | zh | French | fr | | Japanese | ja | German | de | | Korean | ko | Italian | it | | Portuguese | pt | Russian | ru | | Hindi | hi | Arabic | ar | | Cantonese | yue | Turkish | tr | [See full language table in API reference](/voice-cloning/song-generator#supported-languages) *** ## Complete Example ### Reggaeton Track with Manual Lyrics ```json theme={"theme":{"light":"github-light","dark":"github-dark"}} { "key": "your_api_key", "lyrics_generation": false, "lyrics": "[Intro: Sampled Vocal Loop] (Oh-oh-oh-oh-oh-oh-oh-oh) [Chorus] Esta noche todo te lo daré Es libre ya no me amarraré Grita mi nombre, dime que me quieres Me pierdo en tus ojos como si fuera nieve [Verse 1] Tus ojos me hipnotizan, me hacen suspirar Tus labios me llaman, no puedo escapar Tus manos me tocan, siento la pasión Cada latido es una explosión [Chorus] Esta noche todo te lo daré Es libre ya no me amarraré Grita mi nombre, dime que me quieres Me pierdo en tus ojos como si fuera nieve [Bridge - whispered] Solo un instante Deja que te acerque, ven a mí [Final Chorus] Esta noche todo te lo daré Entre tus brazos me quedaré Grita mi nombre, dime que me quieres Me pierdo en tus ojos como si fuera nieve [Outro] Solo una noche más", "caption": "A modern reggaeton track with strong flamenco influence, opening with pitched vocal sample over dembow beat. Clear confident female vocal in Spanish with reverb. Deep sub-bass, crisp drum machine, plucked synth guitar riff. Layered vocals in chorus, atmospheric bridge, sparse whispered outro.", "duration": 199, "language": "es", "webhook": null, "track_id": null } ``` ### Analysis **Caption matches Lyrics:** * ✅ Caption says "reggaeton with flamenco" → Lyrics in Spanish with reggaeton structure * ✅ Caption says "confident female vocal" → Lyrics tone matches * ✅ Caption mentions "whispered outro" → Lyrics has `[Bridge - whispered]` * ✅ Duration 199 seconds appropriate for lyrics amount *** ## Best Practices Summary 1. **Caption First** — Spend time crafting detailed, specific caption 2. **Consistent Description** — Ensure caption and lyrics tell the same story 3. **Calculate Duration** — Count lyrics lines and sections, then estimate time 4. **Use Structure Tags** — Clear sections improve song structure 5. **Test Iterations** — Start simple, then refine based on results 6. **Language Matters** — Set correct language code for best pronunciation *** ## Common Mistakes to Avoid | Mistake | Fix | | - | - | | Too short duration for lyrics | Calculate: lines × 4 seconds + intro/outro | | Conflicting caption and lyrics | Align instruments, energy, vocal style | | Vague caption | Add specific genres, instruments, emotions | | Too many structure tags | Keep tags simple, details in caption | | No section separation | Add blank lines between sections | | Mixed incompatible styles | Either separate or describe as evolution | *** ## Getting Started 1. Start with [Song Generator API Reference](/voice-cloning/song-generator) 2. Try simple examples first 3. Iterate on caption and lyrics 4. Use webhooks for async processing 5. Join our [Discord](https://discord.com/invite/modelslab-1033301189254729748) for community support *** **Need Help?** Check our [API Reference](/voice-cloning/song-generator) or reach out via [Support](https://modelslab.com/support). This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.