Free trial
A free plan grants a one-time quota of 8 songs on the V1 model, limited to four minutes each, with WAV and MP3 downloads and stem extraction listed.
Learn moreFull songs with vocals from lyrics and a style prompt

DiffRhythm is a diss track maker tool. A latent diffusion song generator that turns lyrics plus a style prompt into a complete track with vocals and accompaniment in one pass, capped at four minutes on the free tier and eight on paid plans.
DiffRhythm is a browser-based music generator built on a latent diffusion model. The pitch is deliberately narrow: you supply lyrics and a short style prompt, and the system returns a finished track with both vocals and accompaniment rendered in one pass. There is no separate vocal model, no stem-by-stem assembly, and no multi-stage pipeline to configure. The underlying research comes from the ASLP Lab, and the site is explicit that diffrhythm.ai operates independently of that lab. In practice the workflow is three fields. Lyrics go into a text box, a style descriptor such as pop, rock, ballad, jazz, or drum and bass steers the arrangement, and a generate button starts the render. An instrumental mode skips vocals entirely, and a public display toggle decides whether the result joins the site's showcase gallery. The homepage gallery shows the range this produces: doom metal with bedroom pop touches, Angolan drill, Ukrainian folk-electronic, bluegrass gospel, and zouk percussion, with runtimes from under a minute to over five minutes. The engineering choices behind it are what separate it from older systems. Because generation is non-autoregressive and happens in a compressed latent space rather than on raw audio, the model avoids the drift and slowness that plague long-form sequence generation. The free tier caps output at four minutes on the V1 model, while paid tiers advertise eight-minute songs and access to four model versions. Paid plans also add WAV downloads, stem extraction, vocal removal, and a commercial license. This fits songwriters who have words but no arrangement, and creators who need background tracks in a specific genre fast. It is less suited to producers who want fine control over individual instruments, since style is communicated through prompt text alone. Competitors in this space tend to trade speed for polish or polish for speed; DiffRhythm leans hard toward speed and simplicity, and asks you to judge the output as a demo or sketch rather than a final master.
A free plan grants a one-time quota of 8 songs on the V1 model, limited to four minutes each, with WAV and MP3 downloads and stem extraction listed.
Learn moreA free tier covers 8 songs on the V1 model with four-minute limits. Starter runs $6.99/month billed yearly (30% off $9.90) for 18,000 credits, roughly 12,000 songs. Unlimited is $59/month billed yearly for unlimited generation plus 18,000 bonus credits for stem and vocal tools. Unused credits roll over and plans cancel anytime.
Learn moreSupport is handled by email at support@diffrhythm.ai. Paid tiers list priority support and a priority processing queue, but no documentation portal or help centre is published on the site.
Lyrics plus style prompt input generates complete songs with vocals and accompaniment in a single pass. Non-autoregressive latent diffusion keeps renders fast across genres. Users get instrumental mode, a public gallery toggle, WAV and MP3 downloads, stem extraction, vocal removal, and up to eight concurrent generations on higher plans.
AI Angels400+ customizable AI girlfriends for NSFW chat, voice calls, and roleplay
Candy AIExperience the future of companionship with AI friends who chat, role-play, and connect with you on a human level.
ourdream.aiA vast AI roleplay platform with millions of characters for chat, images, video and voice
AI Angels400+ customizable AI girlfriends for NSFW chat, voice calls, and roleplay
Candy AIExperience the future of companionship with AI friends who chat, role-play, and connect with you on a human level.
ourdream.aiA vast AI roleplay platform with millions of characters for chat, images, video and voice



