Audio Field Guide

The world's most-used end-to-end AI music platform; the V6 family launched September 9, 2026 and is billed by Suno as its biggest upgrade yet. Built in partnership with Warner Music, BMG and Believe using licensed training data, V6 ships in three variants: v6 (flagship, precise control), v6-wild (experimental, more random and bold) and v6-mini (efficient, free for all users). It adds multimodal input (image / video / voice memo to song) and precise local editing (rewrite a section or a single lyric line by natural language, or pull a vocal/drum element from one track into another). Older models will be retired

A professional-grade AI music platform founded by ex-Google DeepMind researchers, renowned for audio fidelity and stem-editing capability. Reached a copyright settlement with Warner Music in Nov 2025, with a jointly launched licensed music platform planned for 2026. Supports fine section control, multi-track separation and post-processing — ideal for professionals who demand high audio quality

An advanced AI music generation model released by Google DeepMind on March 25, 2026 — just one month after Lyria 3. Core breakthrough: single-generation length leaps from 30s to a full 3-minute track, with fine structure customization (intro/verse/chorus/bridge), multimodal input (text/image/video), 24-bit high-quality output, and built-in SynthID watermarking for AI traceability

MiniMax's next-generation music model, API-launched July 16, 2026 and open-sourced August 13, 2026. About 11.1B parameters total (an 8B global LLM for long-range structure and theme consistency, a 0.6B local model for detail, plus a 2.4B Flow-Matching decoder), generating up to 5-minute full songs. Uses 8-layer RVQ hierarchical representation with structured caption control, supporting instrumental mode and lyric structure tags. Weights are released under the Community License with no regional exclusion, deployable locally (~8GB VRAM with layer-streaming)

A full-stack self-developed AI music creation platform by Free Dimension (Shanghai); V3.0 released Feb 12, 2026. Completed a full-pipeline technical rebuild (encode–generate–decode), introducing dual-track modeling and multi-stage reinforcement learning. Singing evolves from 'producing sound' to 'expression' (humming/riffs/breathy voice/natural emotional swings). Guided by 'music equity', it supports text/image/humming multimodal input and outputs full 2–6 minute songs directly — best-in-class Chinese semantic and aesthetic adaptation domestically

The undisputed leader in AI speech synthesis and audio; 2026 ARR surpasses $500M with a $22B valuation (doubled within half a year after Series D). The Eleven v3 model supports 70+ languages and 1000+ preset voices, with industry-top emotional intelligence and context awareness. In 2026 it expanded into AI music generation (ElevenLabs Music), while the ElevenAgents platform handles 33M+ production conversations. Voice cloning takes just 15 seconds

The world's first AI composer registered with a music copyright society (SACEM), focused on purely instrumental scoring. Excels at professional-grade BGM for classical, film, game and advertising scenarios, with MIDI export and fine editing. Generates no vocals — positioned as a professional scoring tool rather than a song-creation platform, with industry-benchmark commercial compliance

A lightweight AI music generation tool based on a WeChat mini program, focused on a zero-barrier experience for domestic users. No app download needed — complete the full flow from lyrics to finished song inside WeChat. Chinese lyric understanding and melody adaptation are tuned to domestic aesthetics; ideal for short-video creators to quickly produce BGM, with better access stability than overseas tools

A domestic integrated AI audio processing tool, known for high-fidelity sound and natural, flowing melodies. Supports precise generation across electronic, pop, ambient and more, with rich detail and clear layering. Integrates the full audio pipeline (generation/editing/mixing) — ideal for content creators who need a one-stop audio solution

Stability AI's next-generation audio model family released in May 2026, in four tiers: small SFX / small (459M params, 2 minutes), medium (1.4B params, 6m20s, open weights) and large (2.7B params, 6m20s, API & paid self-hosting only). The small models target on-device generation while medium handles long audio. Trained on fully licensed data, with partnerships with Warner Music Group and Universal Music Group

End-to-end AI music generation model from Alibaba (Chinese name: 快乐虾米 / 'Happy Shrimp'), launched Aug 17, 2026. Treats music as a special language with grammar, semantics and context; understands natural-language prompts covering style, lyrics, mood, era and vocals, then plans and composes a complete song holistically in one pass. Available on PC web in China and overseas simultaneously, with generous free credits for new users.