Riffusion
AI music generation with web playback
Overview
Riffusion generates music from text prompts using a fine-tuned Stable Diffusion that produces spectrogram images, then converts to audio. Free web player with daily generation limits. Notable as one of the first open-source AI music projects to go viral.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 11 Jul 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
- 3,000 credits/mo (~600 songs)
- 2 concurrent generations
- MP3 downloads
- 10,000 credits/mo (~2,000 songs)
- 8 concurrent generations
- Commercial use
- 30,000 credits/mo (~6,000 songs)
- 12 concurrent generations
- Stem downloads
- 30,000 credits/mo
- 16 concurrent generations
- Early access to new features
- Member events
Use cases
What you can produce with Riffusion
- Generate a complete song from a text prompt describing genre, mood and instrumentation, directly in the browser.
- Create tracks with AI-generated vocals and lyrics, or supply your own lyrics for the model to sing.
- Iterate on a track conversationally, asking the agentic producer to change tempo, swap instruments or rework a section.
- Produce songs up to three minutes long on the post-acquisition ProducerAI platform powered by Google's Lyria 3 and Gemini models.
- Remix or extend an existing generation into alternate versions instead of starting from scratch.
- Download finished tracks as audio files for use in videos, podcasts or personal projects.
- Study or run the original open-source Riffusion spectrogram-diffusion code from its public archive for research and experimentation.
ASEAN Perspective
Riffusion in Southeast Asia
ASEAN-region availability and pricing notes coming soon. Drop the editorial team a note via /contact/ if you can supply local context (Singapore/Malaysia/Indonesia/Thailand/Vietnam).
Riffusion began as a clever diffusion-on-spectrograms experiment and has grown into a full AI music generator that turns text prompts into songs with vocals and instrumentation, notably with a generous free tier. For hobbyists, content creators and quick background tracks it's an accessible, fun way to produce original-sounding music.
It suits creators wanting royalty-friendly music without production skills. Caveats: output quality and control trail leaders like Suno and Udio for consistency and song structure, commercial-use and licensing terms need checking before monetising, and there's no meaningful public API. ASEAN access is global via the web app; lyric/vocal quality is strongest in English.
What people say
Riffusion no longer exists in its original form, and this entry is now largely historical. The project began in 2022 as a viral open-source experiment that generated music by producing spectrogram images with a fine-tuned Stable Diffusion model. It raised about $4 million in seed funding (backers included The Chainsmokers), launched a public beta of a full song-generation service in early 2025 with its FUZZ model family, rebranded to Producer.ai in July 2025 as an "agentic AI music producer", and reached roughly one million monthly users. In February 2026, Google acquired the company, folding the team into Google Labs and DeepMind.
The transition was abrupt for users. Ahead of the February 2026 cutover, users were emailed and told to download their work, because past generations, sessions and media would no longer be accessible once the service switched to Google's models. The relaunched ProducerAI now runs on Google DeepMind's Gemini and Lyria 3 models rather than anything descended from the original Riffusion spectrogram approach, offers tracks up to three minutes, and operates with free and paid plans across 250+ countries.
User sentiment across the project's life had two consistent threads. The original Riffusion earned genuine affection in open-source and hacker communities as a clever, freely available proof that diffusion models could make music, and the 2025 FUZZ-era product was regarded as one of the more credible challengers to Suno and Udio, with free browser-based generation. The Google handover drew criticism for wiping user libraries, a familiar acquisition pattern that stung people who had built up catalogues of generations.
Anyone visiting riffusion.com today is dealing with a Google-owned product, ProducerAI, not the independent startup. It suits hobbyists and creators who want conversational, browser-based music generation on current Google models — but the original open-source Riffusion survives only as an archived codebase.
Summary of public user & expert reviews, compiled by RECATOOLS.
About this listing
This entry was compiled from publicly available data including Riffusion's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with Riffusion unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to Riffusion directly →
Spotted something out of date? Suggest an update →
Riffusion in the news
Alternatives to Riffusion
More in Video & Audio
