> ## Documentation Index
> Fetch the complete documentation index at: https://help.soundverse.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Studio Agent FAQ

## **What is Soundverse Agent?**

**Soundverse Agent** is a conversational “AI producer” that can:

* Use all Soundverse creation tools (Song / Music / Lyrics / Singing / DNA / Melody-to-Song / MIDI-to-Song, etc.).
* Use modification tools (Stem Separator, Inpainting, Extend, Section Analysis, Voice Swap, Voice to Instrument, Similar Music/Song, etc.).
* Analyze key / tempo and change them.
* Apply audio effects (EQ, reverb, delay, compressor, distortion, etc.).
* Layer and merge audio.
* Use system abilities (web search, YouTube URL extractor, record / upload / library selection)
* Maintain context across messages, so follow-ups like “make it a little louder” refer to the last thing you changed.

You always talk to it in natural language. There is no strict syntax.

## **1. General “How do I talk to Agent?” questions**

### **Q1. How do I reference audio?**

You can say:

* “Use **this** track as reference” (after uploading / recording / choosing from library).
* “Take the **song we just generated**…”
* “On the **last track**, make the vocals louder.”
* “Use the **YouTube link I just pasted** as reference.”
* “Use my **DNA model ‘DNA Name’** for style.”

Agent keeps track of the current working song / section unless you switch.

###

### **Q2. How does context memory work?**

Example:

1. “Add reverb to the entire track.”
2. Agent applies reverb.
3. “Increase it a little more.”
   * Agent understands “it” = the reverb you just added.
4. “Now only keep that reverb on the choruses.”
   * Agent uses Section Analysis + effects automation to target choruses.

## **2. Commands by Tool**

Below, for each tool/capability we list **what it does** and **example things you can say**.

### **2.1 Song, Music & Vocal Generation**

#### **2.1.1 AI Song Generator (lyrics → full song with vocals)**

**What it does:** Turns your lyrics (or prompts that include lyrics) into a full song with vocals.

**Things you can say:**

* “Turn these lyrics into a pop song with female vocals in 120 BPM, about 3 minutes long.”
* “Make a dark synthwave song from these lyrics. Keep the chorus big and atmospheric.”
* “Use my lyrics and generate a rock ballad with a long guitar solo at the end.”
* “Generate a Hindi pop song from this verse and chorus, keep it radio-friendly and under 3 minutes.
* “Make a duet version of these lyrics: one male, one female voice, trading lines in the verses.”

You can combine with other constraints:

* “Use my DNA model for the production style, but keep the vocals generic.”
* “Render this in V5 streaming so I can quickly preview it while we edit.”

#### **2.1.2 AI Music Generator (text → music, mostly instrumental)**

**What it does:** Generates music from a text description (instrumental or minimal vocals).

**Things you can say:**

* “Create a 2-minute ambient synth track, 90 BPM, for a sci-fi game menu.”
* “Make a high-energy EDM drop in G minor, 128 BPM.”
* “Generate a chill lo-fi beat with vinyl crackle, slow tempo, for studying.”
* “Give me a cinematic orchestral build-up that resolves after 30 seconds.”
* “Create a soft piano + strings background loop for meditation, around 1 minute and loopable.”

#### **2.1.3 AI Lyrics Writer**

**What it does:** Writes lyrics or rewrites/extends existing lyrics.

**Things you can say:**

* “Write a verse and chorus for a trap song about the football World Cup in Spanish.”
* “Extend these lyrics with a second verse and a bridge, same rhyme scheme.”
* “Rewrite this chorus to be more uplifting but keep the same number of syllables.”
* “Turn this English verse into Hindi, preserving rhyme and rhythm as much as possible.”
* “Give me 10 alternative chorus ideas for this theme: heartbreak but self-empowered.”

You’ll often combine this with **AI Singing Generator** or **Inpainting** to get the audio.

#### **2.1.4 AI Singing Generator**

**What it does:** Takes lyrics (or generated text) and creates realistic singing/spoken vocals.

**Things you can say:**

* “Sing these lyrics in a soft female pop voice, mid-range, no ad-libs.”
* “Generate a rap delivery of these lines in Spanish, keep it tight and rhythmic.”
* “Resing just the chorus in a more powerful belting style.”
* “Make a spoken-word version of this poem over a subtle pad.”
* “Generate backing harmonies for this chorus: one above, one below the lead.”

Often used together with **Stem Separator**, **Voice Swap**, **Inpainting**, and **Section Analysis**.

#### **2.1.5 Melody to Song**

**What it does:** Turns a melody (hummed, sung, or played) into a full musical arrangement.

**Things you can say:**

* “Here’s me humming a melody—turn this into a full synthwave track with drums and bass.”
* “Use this whistled melody as the main hook of a funk song.”
* “Take this melody and build a Bollywood-style orchestration around it.”
* “Turn this melody into a sad piano ballad with minimal percussion.”

You’ll first record or upload the melody.

#### **2.1.6 MIDI to Song**

**What it does:** Uses a MIDI file (chords, melody, progression) and generates a full song.

**Things you can say:**

* “Here’s a MIDI chord progression. Turn it into a deep house track, 122 BPM, with a strong bassline.”
* “Use this MIDI melody as the lead for a trap beat.”
* “Take these chords and make a gospel-inspired arrangement with choir and organ.”
* “Turn this MIDI into a cinematic trailer cue with big drums and strings.”

#### **2.1.7 DNA: Generate & Build**

**What it does:** Uses your DNA models (trained on your songs) to generate in your own style; or builds new DNA from your catalog.

**Things you can say:**

* “Generate 10 song ideas in my ‘BlueNucleus DNA 01’ style, each around 1 minute.”
* “Make a new track using my ‘Lo-fi DNA’ but slower, around 85 BPM.”
* “Train a new DNA model from these 15 tracks I’m uploading called ‘New Album DNA’.”
* “Use my DNA model but make the drums more acoustic and less electronic.”
* “Generate only instrumentals in my DNA style, no vocals.”

You’ll combine with AI Music / Song Generator when specifying more details.

#### **2.1.8 Model Selection (V5 streaming, v4.5 Asian Languages)**

**What it does:** Lets you influence which underlying model/engine Agent uses.

**Things you can say:**

* “Use the V5 model with streaming so I can hear it while it’s generating.”
* “Use the v4.5 model with Hindi/Punjabi support for these lyrics.”
* “Regenerate this but with the higher quality model you used earlier.”

### **2.2 Music Modification, Editing & Analysis**

#### **2.2.1 Stem Separator**

**What it does:** Splits a mixed song into vocal and instrument stems (and possibly multiple instrument stems like drums, guitar, etc., depending on your configuration).

**Things you can say:**

* “Split this track into stems: vocals, drums, bass, other instruments.”
* “Isolate the vocals so we can replace or edit them.”
* “Separate the guitar track so we can add effects only to the guitar.”
* “Give me a version with vocals removed (instrumental).”
* “Give me a version with only drums and bass.”

#### **2.2.2 Extend Music**

**What it does:** Extends a track while keeping style and motifs consistent.

**Things you can say:**

* “Extend this song from 2 minutes to 3:30.”
* “Add a longer intro build-up before the beat drops.”
* “Give me a 30-second outro that feels like a fade-out.”
* “Loop this synth section for another 16 bars before the chorus.”
* “Turn this 15-second idea into a full-length 2-minute track.”

#### **2.2.3 Inpainting (Song & Instrumental)**

**What it does:** Rewrites selected parts of a track, keeping the rest intact.

**Things you can say:**

* “Rewrite just the second half of this verse; keep the rest of the song as-is.”
* “Give me 3 alternate versions of this chorus melody.”
* “Fix this off-key vocal phrase between 0:52 and 0:58.”
* “Replace the drop with a more minimal version, but keep the build the same.”
* “Change the chord progression in the bridge but leave the vocals unchanged.”

You’ll usually specify **time ranges or sections** (e.g., “second verse”, “first chorus”).

#### **2.2.4 Section Analysis**

**What it does:** Detects structural sections like intro, verse, chorus, bridge, drop, etc.

**Things you can say:**

* “Analyze this track and show me the sections: intro, verse, chorus, bridge, outro.”
* “Identify all the choruses so we can edit them.”
* “Tell me where the drops are in this track.”
* “Mark the sections where vocals are present vs instrumental only.”
* “Show me which part could work as a loop for TikTok or Reels.”

Section Analysis is often Step 1 for many edits.

#### **2.2.5 Key & Tempo Detection / Change**

**What it does:** Detects BPM & key, and changes tempo and/or pitch.

**Things you can say:**

* “What is the BPM and key of this track?”
* “Increase the tempo by 10% but keep the pitch the same.”
* “Change the BPM from 100 to 120.”
* “Shift the key up by 2 semitones.”
* “Move this song from C minor to E♭ minor.”
* “Match the tempo and key of this song to that reference track.”

You can also chain instructions:

* “Find the BPM and key of this track, then generate an instrumental in the same BPM and key.”

#### **2.2.6 Effects Engine (Compressor, Reverb, EQ, etc.)**

**AvailableEffects:**

`[`\
`"compressor",`\
`"limiter",`\
`"gain",`\
`"high_pass",`\
`"low_pass",`\
`"peak_eq",`\
`"high_shelf",`\
`"low_shelf",`\
`"reverb",`\
`"delay",`\
`"chorus",`\
`"phaser",`\
`"distortion"`\
`]`

**Things you can say:**

* “Add a bit of reverb to the lead vocal.”
* “Make the whole mix slightly louder, around +1 dB.”
* “Use a high-pass filter on the vocals at 100 Hz to remove rumble.”
* “Warm up the master with a gentle low-shelf boost around 200 Hz.”
* “Make the track sound ‘underwater’.”\
  → Agent might apply **low-pass**, **reverb**, maybe **chorus**
* “Make it sound ‘spacey’ and wide.”\
  → Agent uses **reverb**, **delay**, **chorus/phaser**.
* “Add gentle compression to glue the mix together.”
* “Put a limiter on the master so it doesn’t clip.”
* “Add some distortion to the guitar to make it more aggressive.”
* “Open up the high end of the hi-hats with a high-shelf boost.”

You can then refine with context:

* “Increase the reverb just a little more on the chorus vocals.”
* “Undo the distortion but keep the EQ changes.” (If you implement selective removal.)

#### **2.2.7 Merge Audio (Concatenate)**

**What it does:** Puts clips/tracks one after another.

**Things you can say:**

* “Place this new chorus after the existing one.”
* “Join these three ideas into one track: first A, then B, then C.”
* “Attach this outro after the main song.”
* “After 1 minute, append this instrumental as an extended outro.”

#### **2.2.8 Audio Layering (Stacking / Mixing)**

**What it does:** Stacks tracks on top of each other.

**Things you can say:**

* “Layer this vocal take on top of the main vocal for double-tracking.”
* “Add these new backing vocals under the chorus.”
* “Put this beat under the acapella we extracted.”
* “Blend this pad subtly behind the verses only.”

### **2.3 Controlled Generation Tools**

#### **2.3.1 Voice Swap**

**What it does:** Changes one vocal performance into another voice style.

**Things you can say:**

* “Replace this singer’s voice with my voice model.”
* “Turn this male vocal into a female vocal with a softer tone.”
* “Swap the verse vocals into a gritty rock voice, but keep the chorus as it is.”
* “Use my ‘Creator Voice 01’ for all the leads, and keep backups generic.”

Usually combined with **Stem Separator** and **Section Analysis**.

#### **2.3.2 Voice to Instrument**

**What it does:** Converts a vocal performance into an instrument (e.g., sitar, drums, bansuri, guitar).

**Things you can say:**

* “Turn this hummed melody into a sitar line.”
* “Convert this sung hook into an electric guitar riff.”
* “Use this vocal rhythm to drive a drum pattern.”
* “Turn this vocal line into a bansuri melody layered over the mix.”

#### **2.3.3 Similar Music Generator**

**What it does:** Generates music in the style of a reference track (usually instrumental / production focus).

**Things you can say:**

* “Generate a new synthwave track in the style of this reference track.”
* “Make a similar lo-fi beat but with a different melody.”
* “Create a techno track inspired by this YouTube link, but don’t copy the chords.”
* “Generate 10 background music variations in the style of this game soundtrack.”

You’ll typically supply **reference audio** via upload or YouTube URL extractor.

#### **2.3.4 Similar Song Generator**

**What it does:** Creates full songs (structure, vibe) in a similar style to a reference.

**Things you can say:**

* “Create a song in a similar style to this track, but with new melodies and chord progressions.”
* “Give me 5 variations inspired by this reference but slower and more mellow.”
* “Make a similar Bollywood song but in Punjabi instead of Hindi.”
* “Generate 3 similar songs for this reference that I can use as demos.”

### **2.4 System-Level Tools**

#### **2.4.1 Web Search API**

**What it does:** Fetches up-to-date info from the web.

**Things you can say:**

* “Show me the current trending songs in the world.”
* “What are the top Afrobeats tracks this month?”
* “Find popular synthwave tracks released this year.”
* “What’s trending in Latin trap right now?”

Agent can then use those references to guide style for generation.

#### **2.4.2 YouTube URL Extractor**

**What it does:** Takes a YouTube URL and extracts audio as reference.

**Things you can say:**

* “Use this YouTube song as a reference for vibe: \[paste URL].”
* “Extract the audio from this link and make a similar instrumental.”
* “Take the structure of this song and suggest a similar arrangement for my track.”

#### **2.4.3 Recording, Upload, Library Selection**

**What it does:** Gets audio into the system.

**Things you can say:**

* “Record my voice, I’ll sing a hook.”
* “Use the file I just uploaded as the main track.”
* “Open my library and use the song ‘demo\_05’ as reference.”
* “Record 20 seconds of humming; then turn it into a song.”

## **3. Example Workflows (with Tool Sequences)**

Below are concrete multi-step workflows. For each, we show:

* **User ask**
* **Example prompts**
* **Tool sequence Agent uses**

### **3.1 Trend-based Generation**

#### **Workflow 1 – See global trending songs**

**User ask:** “Show me the current trending songs of the world.”

**Tool sequence:**

1. **Web Search API** → query global charts (e.g., “current top songs worldwide”).
2. Agent summarizes results for the user (titles, artists, styles).

#### **Workflow 2 – Generate 10 synthwave songs in trending style**

**User ask:**\
“In the style of today’s trending songs in synthwave, generate 10 new synthwave songs.”

**Tool sequence:**

1. **Web Search API** → find trending synthwave / related tracks.

2. **YouTube URL Extractor** → optionally pull audio for some of those tracks.

3. For each new song:

   * **Similar Music Generator / Similar Song Generator** → create new synthwave tracks inspired by those references.
   * Optionally **AI Music Generator** → fine-tune prompts (tempo, mood).

4. Present 10 generated synthwave songs.

### **3.2 Direct Prompted Song Creation**

#### **Workflow 3 – Trap song about football World Cup in Spanish**

**User ask:**\
“Make a trap song about football worldcup in Spanish.”

**Example prompts:**

* “Write Spanish trap lyrics about the World Cup, high energy, stadium vibe.”

* “Turn these lyrics into a full trap song.”

**Tool sequence:**

1. **AI Lyrics Writer** → generate Spanish trap lyrics.

2. **AI Song Generator** → create full trap song with Spanish vocals based on those lyrics.

3. (Optional) **Section Analysis** → show user structure.

4. User can then ask for edits (inpainting, effects, etc.).

#### **Workflow 4 – Generate 10 songs in my DNA style**

**User ask:**\
“Generate 10 songs in my DNA style.”

**Tool sequence:**

1. **DNA Generation** → select given DNA model (e.g., “BlueNucleus DNA 01”).

2. For each of 10 tracks:

   * **AI Music Generator / AI Song Generator** with DNA model as style.

3. (Optional) **Section Analysis** → return structures or loops for quick listening.

### **3.3 Voice Replacement & Vocal Editing**

#### **Workflow 5 – Replace song’s vocals with my voice**

**User ask:**\
“Replace this song’s vocals with my voice.”

**Tool sequence:**

1. **Stem Separator** → isolate existing vocals and instruments.

2. **Voice Swap** → apply user’s voice model to the vocal stem; or

   * **AI Singing Generator** → resing using user’s custom voice if supported.

3. **Layering / Merge** → combine new vocal with instrumental.

4. (Optional) **Effects Engine** → match tone (EQ, compression, reverb).

#### **Workflow 6 – Increase volume of vocals in second chorus by 1 dB**

**User ask:**\
“Increase the volume of vocals on second chorus by 1 decibel.”

**Tool sequence:**

1. **Section Analysis** → detect all choruses and locate second chorus region.

2. **Stem Separator** → get vocal stem (if not already isolated).

3. **Effects Engine (gain)** → apply +1 dB gain to vocal stem in second chorus region only.

4. Return result; user may say:

   * “A bit more” → Agent increases gain slightly again, same region.

#### **Workflow 7 – Generate verse with new lyrics alternative**

**User ask:**\
“Generate verse with new lyrics alternative.”

**Tool sequence:**

1. **Section Analysis** → locate the verse you want to replace.

2. **AI Lyrics Writer** → create new verse lyrics (matching theme / syllables).

3. **AI Singing Generator** → sing the new verse.

4. **Inpainting** → replace original verse audio with new verse, keeping surrounding sections intact.

### **3.4 Structural Edits & Looping**

#### **Workflow 8 – Repeat the chorus twice**

**User ask:**\
“Repeat the chorus twice.”

**Tool sequence:**

1. **Section Analysis** → identify chorus region(s).

2. **Merge Audio (concatenate)** → duplicate chorus region and place second copy right after first.

3. (Optional) **Inpainting** → smooth transitions in/out of duplicated chorus.

#### **Workflow 9 – Turn chorus into a perfectly loopable section**

**User ask:**\
“Turn the chorus into a perfectly loopable section.”

**Tool sequence:**

1. **Section Analysis** → isolate chorus region.

2. **Inpainting / Extend Music** → adjust start/end of chorus to loop cleanly (align beats / crossfade).

3. **Export loop** as a stand-alone loop segment.

#### **Workflow 10 – Stitch multiple ideas into one song**

**User ask:**\
“I have three sketches A, B, and C. Join them into one continuous song.”

**Tool sequence:**

1. **Upload / Library selection** → choose clips A, B, C.

2. **Section Analysis (per clip)** → understand structure & key/BPM of each.

3. **Key & Tempo Tools** → adjust clips to a consistent BPM/key if needed.

4. **Merge Audio** → A → B → C in chosen order.

5. **Inpainting** → smooth transitions between each pair.

### **3.5 Tempo, Key & Style Transformations**

#### **Workflow 11 – Identify BPM & key and make instrumental in same style**

**User ask:**\
“Identify the bpm, and key of this track, and make an instrumental track in this style.”

**Tool sequence:**

1. **Key & Tempo Detection** → find BPM and key.

2. (Optional) **Stem Separator** → create instrumental from original (remove vocals).

3. Or **Similar Music Generator** → new instrumental track inspired by reference, using detected BPM/key.

4. Return instrumental to user.

#### **Workflow 12 – Change key and regenerate vocals**

**User ask:**\
“Raise the key of the entire song by 2 semitones and keep the vocals sounding natural.”

**Tool sequence:**

1. **Key & Tempo Detection** → find original key.

2. **Key Change** → plan shift of +2 semitones.

3. **Stem Separator** → isolate vocals.

4. **Inpainting / AI Singing Generator** → re-render vocals in the new key (to avoid artifacts).

5. **Merge Audio** → combine new vocals with key-shifted instrumental.

#### **Workflow 13 – Create a DJ edit at 128 BPM**

**User ask:**\
“Make a DJ-friendly edit of this song at 128 BPM with a longer intro and outro.”

**Tool sequence:**

1. **Key & Tempo Detection** → detect original BPM & key.

2. **Tempo Change** → time-stretch to 128 BPM (keeping pitch).

3. **Section Analysis** → identify intro/outro.

4. **Extend Music** → lengthen intro & outro (e.g., 16/32 bars).

5. **Inpainting** → smooth edits if needed.

### **3.6 Multi-language & Translation Workflows**

#### **Workflow 14 – Translate song into multiple languages and regenerate track**

**User ask:**\
“Translate this song into Hindi, Spanish, Swedish and French. Give me the new track.”

**Tool sequence:**

1. **Stem Separator** → isolate vocals & instrumental.

2. **AI Lyrics Writer** → translate/adapt lyrics into Hindi, Spanish, Swedish, French (matching phrasing).

3. For each language:

   * **AI Singing Generator** → generate vocals in that language.

   * **Layering** → place new vocals over original instrumental.

   * (Optional) **Effects Engine** → match tone & reverb to mix.

4. Return 4 new language versions.

#### **Workflow 15 – Multi-language choruses only**

**User ask:**\
“Make the chorus in Hindi and the verses in English.”

**Tool sequence:**

1. **Section Analysis** → find chorus and verse regions.

2. **AI Lyrics Writer** → translate or rewrite chorus in Hindi; keep verses in English.

3. **AI Singing Generator** → generate Hindi chorus vocals + English verse vocals.

4. **Inpainting** → replace chorus sections only.

5. **Layering / Mix** → combine for final bilingual song.

### **3.7 Detailed Effects & Mixing Automation**

#### **Workflow 16 – Add more reverb to guitar in choruses and lower elsewhere**

**User ask:**\
“Add more reverb to the guitar, but only in all the chorus, decrease the volume of guitar by 20% in rest of the sections.”

**Tool sequence:**

1. **Stem Separator** → isolate guitar stem.

2. **Section Analysis** → detect all chorus sections vs non-chorus sections.

3. **Effects Engine (reverb + gain)**:

   * In chorus sections → increase reverb amount on guitar stem.

   * In all other sections → reduce guitar gain by \~20%.

4. Return new mix.

User can then say things like:

* “The reverb is too much—dial it back slightly.”

* “Make the guitar a bit louder in the bridge only.”

Agent adjusts the same effect chain with new parameters.

#### **Workflow 17 – Make track sound “underwater” or “spacey”**

**User ask:**

* “Make the whole track sound like it’s underwater.”

* “Make the chorus sound more spacey and wide.”

**Tool sequence:**

1. **Effects Engine**:

   * For “underwater” → strong **low\_pass**, extra **reverb**, subtle **chorus**.

   * For “spacey” → **reverb**, **delay**, **chorus/phaser**, slight stereo widening if available.

2. Optionally **Section Analysis** → restrict changes to chorus or target section.

#### **Workflow 18 – Boost only the hook for TikTok**

**User ask:**\
“Make the hook pop more on TikTok: louder vocal, less low-end and more high-end.”

**Tool sequence:**

1. **Section Analysis** → identify hook section(s).

2. **Stem Separator** → isolate vocals & instruments.

3. **Effects Engine**:

   * Increase **gain** on hook vocals.

   * Use **peak\_eq / high\_shelf** to brighten.

   * Possibly **low\_shelf** cut on full mix during hook.

4. Return modified track.

### **3.8 Looping, Short-form & Platform-specific**

#### **Workflow 19 – Create a 1-minute TikTok-friendly version starting from hook**

**User ask:**\
“Make a 1-minute TikTok version of my song that starts directly with the hook.”

**Tool sequence:**

1. **Section Analysis** → find hook/chorus and key sections.

2. **Merge / Arrangement**:

   * Reorder so the song starts with the hook.

   * Trim total duration to \~1 minute.

3. **Inpainting** → smooth transitions / add small intro gap if needed.

#### **Workflow 20 – Make a perfectly loopable 30s background track**

**User ask:**\
“From this song, create a 30-second loopable background segment.”

**Tool sequence:**

1. **Section Analysis** → find sections that can loop well (e.g., a stable groove).

2. **Key & Tempo** → ensure consistent BPM.

3. **Inpainting / Extend Music** → adjust boundaries for clean loop.

4. Deliver the 30-second seamless loop.

### **3.9 Reference-driven Generation**

#### **Workflow 21 – Generate similar beats from YouTube reference**

**User ask:**\
“From this YouTube link, generate 5 similar beats without copying the melody.”

**Tool sequence:**

1. **YouTube URL Extractor** → pull audio.

2. **Section Analysis & Key/Tempo** → understand structure, BPM, key.

3. **Similar Music Generator** → generate 5 different beats inspired by that track’s style.

4. (Optional) **Effects** / **tempo adjustments** as requested.

#### **Workflow 22 – Make a lo-fi version of my song**

**User ask:**\
“Create a lo-fi version of this song for studying.”

**Tool sequence:**

1. **Key & Tempo** → optionally slow the BPM.

2. **Stem Separator** → separate vocals/instruments (if you want instrumental-only).

3. **Similar Music Generator / AI Music Generator** → regenerate with lo-fi vibe (softer drums, vinyl noise, etc.).

4. **Effects Engine** → apply **low\_pass**, gentle **distortion**, maybe **chorus**.

5. Deliver lo-fi mix.

### **3.10 Instrumental & Practice Tracks**

#### **Workflow 23 – Create a karaoke instrumental**

**User ask:**\
“Give me a clean instrumental version for karaoke.”

**Tool sequence:**

1. **Stem Separator** → isolate vocals from instruments.

2. **Mute vocals** → keep only instrumental stems.

3. (Optional) **Effects Engine** → tidy up mix.

4. Output instrumental.

#### **Workflow 24 – Remove drums for practice**

**User ask:**\
“Remove the drums so I can play them live over the track.”

**Tool sequence:**

1. **Stem Separator** → get drum stem (if available) plus rest of instruments.

2. **Mute drum stem**.

3. Output drumless mix.

### **3.11 Game / Batch Content Generation**

#### **Workflow 25 – 20 short game tracks**

**User ask:**\
“Generate 20 short, royalty-free game background tracks, 60–90 seconds each.”

**Tool sequence:**

1. For each of 20 tracks:

   * **AI Music Generator** → with constraints (tempo, mood, length).

   * Optionally, **Similar Music Generator** from a chosen style reference.

2. Return bundle with labeled moods (e.g., menu, battle, victory).

### **3.12 Melody / MIDI-driven Workflows**

#### **Workflow 26 – Turn humming into full song**

**User ask:**\
“I’ll hum an idea; turn it into a full pop song with lyrics and vocals.”

**Tool sequence:**

1. **Record Audio** → capture humming.

2. **Melody to Song** → build instrumental around the melody.

3. **AI Lyrics Writer** → generate lyrics matching melody structure.

4. **AI Singing Generator** → sing the lyrics over the song.

5. (Optional) **Effects** → polish mix.

#### **Workflow 27 – Build EDM track from MIDI**

**User ask:**\
“Here’s a MIDI chord progression—turn it into a big-room EDM track at 128 BPM.”

**Tool sequence:**

1. **Upload MIDI**.

2. **MIDI to Song** → generate arrangement with chosen style.

3. **Key/Tempo** → ensure BPM = 128 and appropriate key.

4. (Optional) **Effects Engine** → extra processing on master.

### **3.13 DNA-Focused Workflows**

#### **Workflow 28 – Train a new DNA model on my catalog**

**User ask:**\
“Train a new DNA model using these 10 songs from my last EP.”

**Tool sequence:**

1. **Upload / Library selection** → choose the 10 songs.

2. **DNA Build / Training** → create new DNA model (e.g., “EP 2025 DNA”).

3. Return status + future prompts:

   * “Now generate 5 ideas using my ‘EP 2025 DNA’.”

#### **Workflow 29 – Batch ideas from DNA**

**User ask:**\
“Generate 20 short ideas in my new DNA style, all around 45 seconds.”

**Tool sequence:**

1. **DNA Generation** → select DNA model.

2. For each idea:

   * **AI Music Generator** → 45-second clips.

3. Collect and present as an idea bank.

### **3.14 Voice Creative Workflows**

#### **Workflow 30 – Voice to instrument lead & harmonies**

**User ask:**\
“Turn my sung melody into a guitar lead, then add harmony lines.”

**Tool sequence:**

1. **Record / Upload vocal**.

2. **Voice to Instrument** → create guitar lead.

3. **Similar Music / AI Music Generator** → generate harmonized variations.

4. **Layering** → stack harmonies on top of original track.

#### **Workflow 31 – Swap my rap verse voice & keep hook original**

**User ask:**\
“Turn my rap verse into a deeper, more aggressive voice, but keep the original hook vocal.”

**Tool sequence:**

1. **Section Analysis** → separate verse vs hook.

2. **Stem Separator** → isolate vocal stem.

3. **Voice Swap** → apply new voice style only to verse regions.

4. **Merge / Layer** → combine swapped verse with original hook.

### **3.15 Structural Feedback & Suggestions**

#### **Workflow 32 – Analyze structure and suggest improvements**

**User ask:**\
“Analyze this track’s structure and suggest changes to make it more catchy.”

**Tool sequence:**

1. **Section Analysis** → map intro, verses, choruses, etc.

2. Agent (LLM) → suggests changes (e.g., “move first chorus earlier”).

3. If user agrees:

   * **Merge / Reorder** sections accordingly.

   * **Inpainting** → smooth transitions.

#### **Workflow 33 – Give me 3 variations of the chorus**

**User ask:**\
“Give me 3 variations of the chorus.”

**Tool sequence:**

1. **Section Analysis** → identify chorus.

2. **Inpainting** → generate 3 alternative versions (melody / production).

3. (Optional) **AI Lyrics Writer + AI Singing Generator** → if user also wants lyrical variations.

4. Provide A/B/C chorus versions.

### **3.16 Batch Hooks & Short Ideas**

#### **Workflow 34 – 20 short hooks about a topic**

**User ask:**\
“Generate 20 different 15-second hooks about space travel in a modern pop style.”

**Tool sequence:**

1. For each idea:

   * **AI Lyrics Writer** → short hook lyrics.

   * **AI Song / Music Generator** → 15-second musical hook.

2. Present 20 hooks as separate snippets.

### **3.17 Complex Example Combining Many Tools**

#### **Workflow 35 – “Identify BPM/key, generate instrumental, edit sections, add effects”**

**User ask (sequence):**

1. “Identify the bpm, and key of this track, and make an instrumental track in this style.”

2. “Give me 3 variation of the chorus.”

3. “Turn the chorus into a perfectly loopable section.”

4. “Add more reverb to the guitar, but only in all the chorus, decrease the volume of guitar by 20% in rest of the sections.”

**Tool sequence:**

1. **Key & Tempo Detection** → BPM, key.

2. **Stem Separator** → create instrumental (mute vocals) or **Similar Music Generator** → new instrumental in same style.

3. **Section Analysis** → locate chorus.

4. **Inpainting** → generate 3 alternate chorus versions.

5. **Extend / Inpainting** → make chosen chorus loopable.

6. **Stem Separator** (if not already) → isolate guitar.

7. **Effects Engine** → chorus: more reverb on guitar; non-chorus: -20% gain on guitar.

8. Deliver final track + loopable chorus segment.

## **4. Contextual Follow-ups – How does “remembering context” work in practice?**

Examples:

* **Effects context**

  1. “Add reverb effect to the entire track.”

  2. *Agent applies reverb.*

  3. “Increase it a little more.”\
     → Agent increases **reverb** level, not something else.

  4. “Now only keep that reverb on the vocals, remove it from everything else.”\
     → Agent uses stems + automation to keep vocal reverb only.

* **Section context**

  1. “Replace the second chorus with a new variation.”

  2. *Agent does it.*

  3. “Actually, also give me a second version that keeps the old chorus but with more drums.”\
     → Agent knows “old chorus” vs “new one,” can branch or create alt take.

* **Reference context**

  1. “Use this YouTube song as reference.” (paste URL)

  2. *Agent extracts and generates similar track.*

  3. “Make another one but slower and with more atmospheric pads.”\
     → Agent keeps using same reference song unless you change it.