If you've searched for an ElevenLabs Music API alternative, you're probably past the "which one sounds best" stage — ElevenLabs Music is a solid text-to-music product on its own. What you're actually solving for is usually one of these:
- You already call hiapi for images or video and don't want a third vendor, a third API key, and a third billing relationship just to add music.
- You need a specific trade-off — longer duration and finer control, or fast turnaround with exact custom lyrics — that doesn't map neatly to a single external tool.
- You want to confirm the output is real, usable audio before you commit a workflow to it, not just read a spec sheet.
This piece covers the second and third points directly with hiapi's two live music models: minimax-music-3 and lyria-3.5. We ran one real generation on each — not a mockup, not a spec comparison — so you can hear the actual output quality before deciding.
Two Models, Two Different Jobs
hiapi doesn't have a single "music" model — it has two, and they're built for different requests:
minimax-music-3 generates a complete track (instrumental or with vocals) up to 5 minutes long, at 44.1kHz stereo WAV. Duration, mood, instrumentation, and vocal style are all steerable through the prompt and optional lyrics field. It's priced per second of output ($0.0026/s — a 30-second clip runs about $0.08), so cost scales with how long a track you actually need.
lyria-3.5 is built around custom lyrics. You supply the words, a style/mood description, and it returns a finished song for a flat $0.09 per song, regardless of length. There's no duration dial to turn — you're paying for one complete generation.
The practical split: if you need a specific-length instrumental bed with tight control over exactly how long it runs, minimax-music-3 is the better fit. If you have actual lyrics you want set to music and don't want to think about duration or pricing math, lyria-3.5 gets there faster.
Real Test #1: Instrumental Background Track (minimax-music-3)
We asked for a 30-second instrumental bed — the kind of thing you'd drop under a product demo video.
Prompt used (verbatim, no lyrics — instrumental only):
Upbeat corporate tech background music, confident and modern, driving
mid-tempo synth pulse with a clean plucked arpeggio, light electronic
percussion, optimistic and professional mood, suitable for a product
demo video
Result: Listen to the minimax-music-3 sample (WAV, 30s)
The output came back as a genuine 44.1kHz stereo WAV file (verified — not a compressed preview), matching the model's advertised spec exactly. The arpeggio and percussion layering followed the prompt closely, and the track loops cleanly enough to extend for a longer video without an obvious seam.
Real Test #2: Custom-Lyrics Vocal Track (lyria-3.5)
For lyria-3.5, we wrote actual lyrics rather than a mood description alone — this is the model's core differentiator.
Prompt used:
Warm indie pop, bright acoustic guitar strums, soft claps, hopeful
female vocal, easy driving tempo
Lyrics supplied:
Verse: Small steps forward, the light is turning gold
Chorus: We keep the melody, we make it ours to hold
Result: Listen to the lyria-3.5 sample (30s)
The vocal delivery tracked the supplied lyric lines rather than generating placeholder syllables, and the guitar/clap instrumentation matched the requested style. This is the clearest way to evaluate lyria-3.5 for your own use case: if you already have lyrics (jingle copy, a verse for a brand campaign, a short intro song), this is a direct text-in, song-out path.
Same API, No New Vendor
Both models run through hiapi's standard async task endpoint — the same POST /v1/tasks → poll GET /v1/tasks/<id> pattern used for image and video generation on the platform. If you're already integrated with hiapi for visuals, adding music is a new model value and a different input shape, not a new integration.
That also means one API key, one usage dashboard, and one invoice — which is the practical answer to "why not just use ElevenLabs Music directly": if ElevenLabs isn't already in your stack, adding it means a second account, a second key to rotate, and a second bill to reconcile against usage. If hiapi already is in your stack, it isn't.
Where the Limits Are
Neither model is a full production-music suite:
- No stem export. Both return a single mixed audio file — no separated vocal/instrumental tracks.
- No lyric-to-timestamp sync. If you need word-level timing for captions or karaoke-style displays, that's not part of either model's output today.
- minimax-music-3's lyrics field is optional but not a fine-grained vocal director — for precise word-for-word vocal delivery, lyria-3.5's lyrics-first design is the closer fit of the two.
If your workflow needs any of the above, check the model pages linked below for current capabilities before committing — model capabilities on the platform do get extended over time.
FAQ
Is there a free tier to test these models? Pricing is usage-based per generation ($0.0026/second for minimax-music-3, a flat $0.09/song for lyria-3.5) — there's no separate free quota beyond whatever trial credit your hiapi account starts with. Check hiapi's pricing page for current account-level details.
Can I generate a full 5-minute song with minimax-music-3? Yes — 5 minutes is the model's maximum duration. At $0.0026/second, a full 5-minute track costs roughly $0.78.
Does lyria-3.5 support languages other than English? The lyrics field accepts the text you provide directly; delivery quality for non-English lyrics will depend on the specific language and isn't something we tested in this piece — verify with your own lyrics before committing to a production workflow.
Which model is closer to ElevenLabs Music's workflow? lyria-3.5's lyrics-in, song-out flow is the more direct match if you're coming from a lyrics-driven workflow. minimax-music-3 is the better fit if you're used to specifying duration and mood and want instrumental control.
Can I use these tracks commercially? Usage rights follow hiapi's standard terms of service for generated output — review the terms for your account tier before using generated tracks in commercial releases.
Takeaways
- minimax-music-3 and lyria-3.5 are both live, real, self-run models on hiapi — not roadmap items.
- minimax-music-3 fits duration-controlled instrumental or vocal tracks up to 5 minutes, billed per second.
- lyria-3.5 fits lyrics-first vocal tracks, billed as a flat $0.09/song regardless of length.
- Both share hiapi's existing task API, so adding music to an existing image/video integration doesn't require a new vendor relationship.
- Neither model currently offers stem separation or lyric timestamp sync — confirm current capabilities on the model pages if your workflow needs them.
Explore the full specs and try it yourself on the minimax-music-3 model page or the lyria-3.5 model page. For a deeper API walkthrough, see How to Use the minimax-music-3 API and Lyria 3.5 for E-commerce Music.
