SonicBliss (under Produce → SonicBliss) turns a text prompt into production-ready audio. Describe what you want to hear, pick a mode and a duration, and the track lands in your Library as a normal audio asset — with a waveform poster, a license badge, and full generation provenance.
Gating: Creator and up (sonicbliss entitlement). Every generation spends credits from your plan’s monthly budget.

​
The four modes

Sound FX

One-shots, foley & ambience — a UI click, a door slam, rain on a tin roof, a character sound. Short, precise, ready to drop into a cut.

Instrumental

Beds, beats & scores — background music with no vocals. Steer it with genre, mood, and instrument chips (“warm lo-fi hip-hop, dusty Rhodes, vinyl crackle”).

Full song

Sung vocals from your lyrics. Paste lyrics (or let the AI lyric-writer draft them) and describe the genre, voice, and mood.

Loop

Seamless beds & grooves that cycle cleanly — ideal under looping video, ads, and social clips.

​
Composing a prompt

  • Preset chips — each mode offers starter chips (type, genre, mood, instruments, voice) that build a prompt for you; the dice button drops in a random starter prompt.
  • Prompt library — three layers in one panel: operator starters curated for your instance (a large, categorized starter set spanning every mode and genre), your saved prompts, and favorites. Save anything you like, search across all of it, and reuse with one click.
  • Model picker — the model dropdown is operator-curated per mode; the default is picked for you, so you only need to touch it when you want a specific engine.

​
Options

Hit Generate — the job streams progress live in the Activity dock, and each variation appears in the tracks grid as it finishes.

​
The AI lyric-writer

In Full song mode, click the lyric-writer, describe a theme, genre, mood, and optional structure (verse/chorus/bridge), and it drafts complete lyrics straight into the lyrics box — then edit them like any text before you generate.

​
Match a video

Scoring to picture? Click Match a video to open the media picker — search your Library videos by name, filter by project, sort by newest/name/length, and preview any clip full-screen before choosing (double-click or double-tap to pick). The duration snaps to that video’s exact length, so the generated bed fits your cut with no trimming.

​
Long tracks, consistent loudness

  • Auto-stitching — most models cap out well under 10 minutes. Ask for a longer duration and SonicBliss generates in segments and stitches them with crossfades automatically; you just get one continuous track.
  • Loudness-normalised — every output is normalised to -14 LUFS (the streaming standard), so generated tracks sit at a consistent level next to each other and your other media.

​
Your tracks

Results land in the Library as audio assets and in the SonicBliss tracks grid:
  • Hover a card to preview it; double-click (double-tap or the expand button on touch) opens a popup player.
  • Every track carries a license/source badge and full provenance (gen_meta): the model that made it, its license and commercial-use status, plus the prompt and seed — so you can always answer “where did this audio come from, and can I use it commercially?”
  • One click hands the track off to the VEditor to drop it into a cut.

​
Stems & extend

Two follow-up actions on any generated track:
  • Split into stems — separates the track into vocals / drums / bass / other; each stem lands in your Library as its own audio asset.
  • Extend — adds 5–300 s of new material onto the end of the track, optionally with a fresh prompt to steer where it goes.
Both run as jobs (watch them in Activity) and both spend credits.

​
Credits

Every generation spends from a monthly credits budget (1 credit = $0.01): Audio costs ≈0.5 credits per second generated (video in ClipForge ≈8/s); operators can tune the rate per model. Run out mid-month? Top up with credit packs — no plan change needed.

​
Troubleshooting

  • “Model busy” — the provider is under load; wait a moment and retry (or pick a different model from the dropdown).
  • “No key” / provider not configured — the instance operator hasn’t configured that model’s provider yet. Pick another model, or contact support so an operator can enable it.
  • Job seems stuck — open Activity; jobs stream live progress, and a failed job shows the reason there.

​
Over the API

Everything above is scriptable — the routes appear in the API Reference tab under the sonicbliss and genai tags:
  • POST /sonicbliss/generate → 202 {job_id} — mode, prompt, lyrics, duration_seconds (1–600), format, model_key?, variations (1–4), seed?, bpm?, key_sig?, project_id?.
  • GET /sonicbliss/tracks — list generated tracks.
  • POST /sonicbliss/{asset_id}/stems → 202 — split into vocals/drums/bass/other.
  • POST /sonicbliss/{asset_id}/extend → 202 — add_seconds (5–300), optional prompt.
  • POST /sonicbliss/lyrics — theme/genre/mood/structure → {lyrics}.
  • GET /genai/models?feature=&mode= · /genai/prompts (+ PUT /genai/prompts/prefs) — models and the prompt library.
Follow any job_id on GET /jobs/{id}/events (SSE) — see the Developer guide for a full recipe.