Bygheart Music — Songwriting Guide

Models & credits

Three DiT models live on the GPU at all times. Composer Lab piggybacks on Studio Pro until the 4th handler ships.

Pro plan: 200 credits/month, 30/day cap, up to 60 unused credits roll over. Failed generations are refunded automatically. Ultimate: unlimited.

Modes

Song Description (prompt)

The single most important control. Pack it with what you actually want to hear: genre, mood, vocal style, instruments, mix quality, hook size, energy. Avoid generic words.

epic modern pop, dark dance pop, cinematic trap-pop, stadium anthem, huge hook, distorted bass, choir layers, glossy radio production, powerful female lead vocal, stacked harmonies, explosive chorus, dramatic bridge, punchy drums, wide synths, polished high-energy mix

pop song, female vocals

Pro tip: write the description like you’re briefing a producer. Specifics beat adjectives.

Negative styles

Comma-separated list of what to avoid. Set it in Advanced Options → Negative styles.

country, acoustic folk, screamo, metalcore, lo-fi bedroom pop, muddy mix, weak vocals, thin drums, pitch wobble, wrong language

A good default if you’re going for radio-quality pop: muddy mix, weak vocals, thin drums, gibberish.

Lyrics structure

Lyrics are the song’s timeline. Section tags drive arrangement. Keep production/mood mostly in the prompt; reserve in-line vibe hints for transition moments.

Section tags the model understands

Add a short hint in the tag for a vibe shift: [Bridge - stripped, emotional] or [Final Chorus - stadium ending].

Recommended template

[Intro]

[Verse 1]

[Pre-Chorus]

[Chorus]

[Verse 2]

[Pre-Chorus]

[Chorus]

[Bridge]

[Final Chorus]

[Outro]

Vocal style

ACE-Step has no hard gender or style switch. The Studio’s Voice + Delivery dropdowns inject phrases into your prompt and lyric tags so the model is biased toward that style. They’re strong suggestions, not guarantees — back them up in the Song Description.

Voice

Delivery

In-lyric performance tags

Drop these inline in your lyrics or tag a section with them. They tell the model how to perform that line/section.

Duration, BPM, Key, Time signature

Expert controls (Advanced Options)

Defaults work fine. These are for fine-tuning when you know what you’re doing.

Edit / Repaint mode

Upload a song and rewrite parts of it (lyrics or section). The repaint engine generates new audio for masked sections and crossfades it back into the source.

Remix / Cover mode

Re-imagine a song in a new style. Upload a reference and write a Song Description for the target vibe.

Full examples

A. Modern pop anthem (Custom mode, Studio Pro)

epic modern pop, dark dance pop, cinematic trap-pop, stadium anthem, huge hook, distorted bass, choir layers, glossy radio production, powerful female lead vocal, stacked harmonies, explosive chorus, dramatic bridge, punchy drums, wide synths, polished high-energy mix

[Intro - dark synth pulse]

Midnight
Gold light
Black car
White lies

[Verse 1 - tight modern pop flow]

I got a new face in the rearview
Old me tried to call, I hit decline
Heart in my chest like a loaded weapon
Aiming at the version I left behind

[Pre-Chorus - rising]

Tell the angels I’m unavailable
Tell my demons they can wait in line
Got a thousand voices in my head tonight
But the only one I’m hearing is mine

[Chorus - huge, anthemic]

I go god mode after midnight
Hands up when the lightning hits
Diamonds in the dirt, baby that’s my life
I’m the storm, I’m the calm, I’m the lit fuse fire

[Bridge - stripped, emotional]

I used to beg for a little mercy
Now mercy looks like me
I used to fold like paper
Now I burn the whole damn page

[Final Chorus - massive stadium ending]

I go god mode after midnight
Whole world watching me ignite
Every scar a spotlight, every tear a strobe
I’m the storm, I’m the calm, I’m the lit fuse fire

[Outro - choir and vocal chop fade]

Midnight
Gold light
Black car
White lies

B. Lo-fi instrumental (Simple mode, Fast)

C. Spoken-word rap (Custom, Studio Max)

Troubleshooting

API reference

Pro+ users can generate songs programmatically. Get a key at /api-keys.html. Quota is shared with the Studio. Failed generations are auto-refunded.

POST /bygheartsong/api/release_task

curl -X POST https://iamvision.tech/bygheartsong/api/release_task \
  -H "Authorization: Bearer bgh_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "bygheart-studio-pro",
    "prompt": "epic modern pop, female belt vocal, glossy mix",
    "lyrics": "[Verse]\n...\n[Chorus]\n...",
    "audio_duration": 180,
    "audio_format": "mp3",
    "vocal_language": "en",
    "use_random_seed": true,
    "task_type": "text2music",
    "lm_negative_prompt": "muddy mix, weak vocals"
  }'

All accepted fields

Bold = required-ish for a useful song. Italics = advanced.

POST /bygheartsong/api/query_result

curl -X POST https://iamvision.tech/bygheartsong/api/query_result \
  -H "Authorization: Bearer bgh_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{ "task_id_list": ["TASK_ID_FROM_RELEASE"] }'

GET /bygheartsong/api/v1/audio?path=...

curl -L -H "Authorization: Bearer bgh_YOUR_KEY" \
  "https://iamvision.tech/bygheartsong/api/v1/audio?path=/abs/path/from/result.mp3" \
  -o song.mp3

Errors