Show HN: JJ Suna AI – Open 5B Flow-Matching DiT for 48 KHz Music Synthesis

JJ Coders Group released Suna AI v1.0, an open 5-billion-parameter flow-matching diffusion transformer that synthesizes 48 kHz stereo music, publishing full model weights, the producer GGUF brain, DNA scale profiles and a live inference API with no gated paywalls. The pipeline combines an ACE-Step 1.5 XL SFT continuous flow-matching diffusion core with a 1.5B Qwen2.5-Instruct producer model and 14 Suna Rank-64 LoRA experts, and exports uncompressed 48 kHz 24-bit PCM or 320 kbps MP3 through three-stage mastering. Outputs are distributed under the Suna Community License, which the company says lets creators monetize original compositions across streaming, games, broadcast TV, podcasts and digital media with zero ongoing royalty dues.

Where human artistry meets acoustic intelligence. Suna AI v1.0 delivers 48 kHz True Stereo Broadcast Audio synthesized through our 5-Billion Parameter Diffusion Pipeline with native Sing-to-Masterpiece Engine . OPEN FOUNDATION ARCHITECTURE & COMMUNITY BENCHMARKS Experience real songs synthesized end-to-end on Suna AI's 5-Billion Parameter Diffusion Pipeline . Click any track to stream instantly in lossless 48 kHz stereo. Atmospheric Piano & Dynamic Stadium Pop-Rock Drop Solemn atmospheric piano building into an explosive, soaring pop-rock stadium chorus with live acoustic drums. Infectious Afro-house rhythm, acoustic nylon guitar riffs, syncopated festival percussion, sub-bass, and festive brass. High-energy stadium rock synthesis with instant kick and sub-bass rhythm lock, electric guitar riffs, and studio dynamics. Brooding staccato cello and violin strings, heroic brass fanfare, rolling trap hi-hats, deep 808 sub-bass glides, and vocal delivery. Suna AI v1.0 bridges the divide between open foundation weights and commercial studio workflow. Every waveform is rendered with true 48.0kHz clarity, dual-brain coordination, and zero algorithmic slop. Combines an ACE-Step 1.5 XL SFT continuous flow-matching diffusion core with a 1.5B Qwen2.5-Instruct Producer brain and 14 Suna Rank-64 LoRA experts. The producer orchestrates harmonic arrangement while the continuous flow-matching diffusion network renders true 48 kHz stereo broadcast waveforms. Export in uncompressed 48 kHz 24-bit PCM / 320 kbps MP3 with 3-stage studio mastering: sub-bass contouring 40–70 Hz , transient punch, and true-peak stereo Haas widening. Generated outputs are distributed under the permissive Suna Community License. Creators retain rights to monetize their original compositions across streaming, video games, broadcast TV, podcasts, and digital media with zero ongoing royalty dues. Full model weights, producer GGUF brain, DNA scale profiles, and live inference API are available to all developers and researchers. Zero gated paywalls. Every creation is guided by a four-step pipeline: from prompt vector analysis, arrangement, waveform synthesis, and final broadcast mastering. Natural language acoustic parsing analyzes rhythmic cadence, emotional sentiment, and reference hit artists, matching optimal BPM, musical key, and instrumentation tags. // Step 1: Prompt Intent & Acoustic Vector Mapping user input = { prompt: "cinematic stadium rock anthem with live acoustic drums" } // Matched: "Atmospheric Stadium Rock" - 170 BPM / C Major // Energy Level: High • Brightness Contour: +2dB EQ Scientifically calibrated spectral curves, dynamic transient density, and RMS sub-bass harmonic profiles engineered natively by Suna AI's neural synthesis models. Official governance for suna.jjcoders.com & jjcoders.com Last updated: October 2026 • JJ coders Group of Companies THE SUNA AI SERVICE, FOUNDATION WEIGHTS jjcoders/suna-v1 , SYNTHESIZED AUDIO STEMS, AND WAVEFORMS ARE PROVIDED STRICTLY ON AN "AS IS" AND "AS AVAILABLE" BASIS, WITHOUT WARRANTIES OF ANY KIND, WHETHER EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE, TITLE, OR NON-INFRINGEMENT. JJ CODERS GROUP MAKES NO WARRANTY OR REPRESENTATION REGARDING UNINTENTIONAL HARMONIC, TIMBRAL, OR MELODIC SIMILARITIES TO PRE-EXISTING COMPOSITIONS THAT MAY ARISE THROUGH STATISTICAL NEURAL DIFFUSION. TO THE MAXIMUM EXTENT PERMITTED BY APPLICABLE JURISDICTIONAL LAW, JJ CODERS GROUP, ITS DIRECTORS, EMPLOYEES, AFFILIATES, AND AGENTS SHALL NOT BE LIABLE FOR ANY INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, PUNITIVE, OR CONSEQUENTIAL DAMAGES, NOR FOR ANY LOSS OF REVENUE, PROFITS, OR COMMERCIAL OPPORTUNITY ARISING OUT OF OR IN CONNECTION WITH THE ACCESS, USE, OR INABILITY TO USE GENERATED AUDIO OUTPUTS. TOTAL AGGREGATE LIABILITY OF JJ CODERS GROUP UNDER ALL CAUSES OF ACTION SHALL BE LIMITED TO THE TOTAL FEES PAID BY YOU IN THE PRECEDING TWELVE 12 MONTHS, OR ONE HUNDRED U.S. DOLLARS $100.00 USD , WHICHEVER IS LESS. Users explicitly agree not to submit prompts, tags, or vocal stems designed with the intent of producing counterfeit reproductions or infringing upon trademarks, artist likenesses, or copyrighted master recordings. Users agree to defend, indemnify, and hold harmless JJ coders Group against any claims, damages, or liabilities resulting from user's custom prompts or unauthorized distribution of third-party protected elements. Effective Date: October 2026 • JJ coders Group European Union and UK residents maintain enforceable rights to request access, rectification, portability, and permanent erasure "right to be forgotten" of personal data. Suna AI collects minimal telemetry necessary for server stability and does not sell or lease personal data to commercial brokers. We do not "sell" or "share" personal information or audio assets for cross-context behavioral advertising. For statutory verification or data requests, submit requests to: privacy@jjcoders.com. Effective Date: October 2026 suna.jjcoders.com operates exclusively with lightweight, essential client-side storage keys in your browser's LocalStorage. These are deployed solely to preserve user playback volume, theme preferences, and cookie consent status suna cookie accepted . We deploy zero third-party behavioral trackers, zero marketing analytics scripts, and zero advertising beacons. Statutory Notice & Counter-Notification Procedure JJ coders Group respects the intellectual property rights of creators and adheres strictly to the Digital Millennium Copyright Act 17 U.S.C. § 512 . If you believe in good faith that any content synthesized or presented on suna.jjcoders.com infringes upon your copyright, please deliver a formal Notice of Infringement to our Designated Copyright Agent: Under 17 U.S.C. § 512 g , any user who believes an output was improperly disabled or misidentified may file a statutory Counter-Notice. JJ coders Group maintains an expeditious termination policy for repeat infringers who intentionally abuse the service to violate third-party rights. Tennessee ELVIS Act 2024 , BIPA 740 ILCS 14/ & COPPA Compliance Under the Tennessee Ensuring Likeness Voice and Image Security ELVIS Act of 2024 and statutory right-of-publicity protections, an individual's voice is a protected personal right. Users of the Sing-to-Masterpiece Engine warrant and represent that they own or hold express written licensing rights to any vocal audio uploaded. The synthesis of unauthorized celebrity vocal clones, deceptive soundalikes, or non-consensual voice deepfakes is strictly prohibited. Suna AI does NOT collect, extract, capture, or store biometric voiceprints or vocal identifier templates. Acoustic conditioning algorithms evaluate raw frequency, pitch trajectory, and envelope transients in volatile memory solely for dynamic MIDI and harmony alignment, and all representations are discarded immediately upon stream completion. Suna AI is designed for professional creators, artists, and general audiences aged 13 and older. We do not knowingly collect, retain, or solicit vocal recordings or personal identifiers from children under 13. Inquiries: coppa@jjcoders.com.