MusicGen (MG-Medium)
Free.ai
·
music
·
~500 tokens per track
·
4.1
from 11 users of this category
MusicGen (MG-Medium) is a music generation model. Routed through external models — ~500 tokens per track (50% markup over upstream cost).
Use via API
OpenAI-compatible REST API. Generate a key and call this model in seconds.
curl -X POST https://api.free.ai/v1/music/generate/ \
-H "Authorization: Bearer sk-free-..." \
-H "Content-Type: application/json" \
-d '{"model":"premium/musicgen","prompt":"your prompt here"}'Similar models
Frequently Asked Questions
MusicGen (MG-Medium) generates original instrumental music (and on some engines, vocals) from a text prompt. Describe the mood, genre, instrumentation, or BPM and MusicGen (MG-Medium) returns a finished audio track.
10 to 60 seconds per generation depending on the engine. For longer pieces, chain multiple generations through /music/stitch/ or generate in 30-second segments and crossfade.
Some engines synthesize vocals from lyrics you provide (paste lyrics in the form); others are instrumental-only. Check the form on this page - a Lyrics field means vocals are supported.
Pop, rock, EDM, jazz, ambient, classical, lo-fi, hip-hop instrumentals all render reasonably. Niche genres (gamelan, Carnatic, throat singing) are weaker - describe instruments explicitly for best results.
MusicGen (MG-Medium) is a premium music model. Each track runs ~5,000-15,000 tokens depending on duration and complexity, shown live before you generate. Self-hosted music runs free on your 30,000-token daily pool; premium models are pay-as-you-go, with top-ups from $1.
WAV by default (lossless, ready for further mixing). MP3 is available in the output picker for smaller files.
No. MusicGen (MG-Medium) returns a mixed track. To split it into stems, run the output through /music/stem-splitter/, which extracts vocals, drums, bass and other.
Yes - /batch/ accepts a list of prompts. Each track lands in /account/?tab=history with the prompt as the filename. For more control, use the API.
Yes - POST your prompt to /v1/music/generate/ with model="MusicGen (MG-Medium)". /api/ has the full reference including lyrics, duration, and tempo parameters.
MusicGen (MG-Medium) does not reproduce copyrighted compositions - outputs are synthesized fresh. Free.ai grants commercial use; you keep the rights to your generated tracks. Naming an artist in the prompt asks for "in the style of" - the actual audio is original.
15 seconds to 2 minutes per track depending on duration + engine. Premium models finish faster. Use the queue button to close the tab - result lands in /account/?tab=history when done.
Yes - failed renders auto-refund. For tracks that ran but sound bad, email contact@free.ai with the share link and we will review case-by-case.