Short-form video and ad creative live and die on pacing, and pacing needs music. Licensing stock tracks per-clip doesn't scale when you're producing dozens of variants a week, and generic royalty-free libraries rarely fit the exact mood or length you need. minimax-music-2.6's Instrumental Mode solves this with a single API call: describe the mood and instrumentation you want, skip the lyrics, and get back a complete, loop-friendly instrumental track built for background use.
What Instrumental Mode Actually Does
minimax-music-2.6 is a task-based text-to-music model. There's no separate "instrumental" boolean anywhere in the schema — omitting the optional lyrics field is what triggers Instrumental Mode. Send lyrics and you get a full vocal song; leave it out and you get a clean instrumental built from your prompt alone.
| Field | Required | Notes |
|---|---|---|
prompt | Yes | Style, instrumentation, and mood description |
lyrics | No | Omit entirely for Instrumental Mode |
sample_rate | No | Enum: 16000 / 24000 / 32000 / 44100 |
bitrate | No | Enum: 32000 / 64000 / 128000 / 256000 |
There's no duration, seed, or reference-audio parameter — the model decides song length and structure on its own from the prompt.
The Exact Prompt We Ran
Upbeat lo-fi hip-hop instrumental for short-form video and ad background, warm electric piano chords, mellow boom-bap drums, soft vinyl crackle, light bass groove, no vocals, loopable, calm and focused mood
🎧 Listen to the generated instrumental track (mp3)
That's the complete, unedited prompt — no vocals requested, no lyrics field sent, and the result came back as a full instrumental cue ready to drop under a video timeline.
Calling the API: Task Submit + Poll
minimax-music-2.6 runs through hiapi's unified async task endpoint, like every other task-based model on the platform:
curl -X POST https://api.hiapi.ai/v1/tasks \
-H "Authorization: Bearer $HIAPI_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "minimax-music-2.6",
"input": {
"prompt": "Upbeat lo-fi hip-hop instrumental for short-form video and ad background, warm electric piano chords, mellow boom-bap drums, soft vinyl crackle, light bass groove, no vocals, loopable, calm and focused mood"
}
}'
Then poll the task until it reaches a terminal state and read the output URL:
import time
import requests
headers = {"Authorization": f"Bearer {HIAPI_API_KEY}"}
resp = requests.post(
"https://api.hiapi.ai/v1/tasks",
headers=headers,
json={
"model": "minimax-music-2.6",
"input": {"prompt": "Upbeat lo-fi hip-hop instrumental for short-form video and ad background, warm electric piano chords, mellow boom-bap drums, soft vinyl crackle, light bass groove, no vocals, loopable, calm and focused mood"},
},
)
task_id = resp.json()["id"]
while True:
task = requests.get(f"https://api.hiapi.ai/v1/tasks/{task_id}", headers=headers).json()
if task["status"] in ("succeeded", "failed"):
break
time.sleep(3)
audio_url = task["output"][0]["url"]
Output URLs are time-limited, so download the bytes (or re-host them) as soon as the task succeeds rather than linking the raw response URL.
Pricing: 2.6 vs. the Rest of the MiniMax Music Family
minimax-music-2.6 bills a flat $0.21 per song, no matter how long the generated track ends up being. That flat-rate model matters for background-music batches, where you don't want cost to spike just because one track came back longer than expected.
- minimax-music-2.6 — $0.21/song flat, vocals or Instrumental Mode, auto lyrics and song-structure tags supported.
- minimax-music-1.5 — $0.07/song flat, natural vocals only, tracks run up to roughly 4 minutes. Cheaper, but no instrumental-only path.
- minimax-music-3 — $0.0026 per second of output (usage-based), tracks up to 5 minutes at 44.1kHz stereo WAV. Cheaper for short clips, more expensive than 2.6 once a track runs past a couple of minutes.
If your use case is short (15-60 second) background beds for reels, TikToks, or pre-roll ads, 2.6's flat per-song price is easier to forecast than pure per-second billing.
Where This Fits in a Short-Form Video Pipeline
Background music is one piece of an AI-produced short-form video stack. On the video side, our short-form video workflow with Seedance 2.0 covers generating the actual clips; minimax-music-2.6 fills in the audio bed once the visuals are locked. If you're producing audio for retail or product placements instead of short-form video specifically, using Lyria 3.5 for e-commerce and retail music is a useful comparison for a different use case with different mood requirements.
Cover Art
Clean modern illustration for a tech blog cover: a smartphone screen showing a vertical short-form video timeline with a colorful audio waveform track underneath it, minimal flat design, soft orange and dark navy color palette, abstract music note icons floating around, no readable text or logos, wide 16:9 composition
FAQ
Does minimax-music-2.6 support lyrics, or is it instrumental-only?
Both. Send a lyrics field and you get a full vocal song; omit it entirely and the model generates an instrumental track instead. There's no separate mode flag — leaving the field out is the trigger.
Does a longer track cost more? No. minimax-music-2.6 bills a flat $0.21 per generated song regardless of length. That's different from minimax-music-3, which bills $0.0026 per second of output.
Can I control the audio format?
Yes, within limits. sample_rate (16000/24000/32000/44100) and bitrate (32000/64000/128000/256000) are both optional enums. There's no duration, seed, or reference-audio parameter to steer song structure beyond your prompt.
What's the difference between minimax-music-2.6 and minimax-music-1.5? minimax-music-1.5 is cheaper ($0.07/song) but vocal-only, with tracks running up to about 4 minutes. minimax-music-2.6 adds Instrumental Mode, automatic lyrics generation, and song-structure tags for $0.21/song.
Try It
If you need royalty-free background music for short-form video or ad creative on a predictable per-track budget, try minimax-music-2.6 with your own prompt.
