最新资讯
dev.to 精简 MCP server 工具数量:更少工具、更小影响面,双重收益 dev.to 为什么 git checkout 无法撤销 AI 编码 agent 的改动? wechat AI短剧门槛降降降!实测即梦短剧工作室:剧本分镜预演一条龙,伙伴计划长期护航 dev.to 让模型改写文档前,先量化事实保真度 dev.to 回放 153 个 Claude Code 与 Codex 会话,审计 Agent 的生产影响范围 dev.to 银行 77 意图分类最便宜 LLM:GPT-6.1 Terra 每次成功任务 $0.000727 theinformation.com Chesky的副业:Airbnb向CEO新设AI实验室出售IP theverge.com OpenAI 宣布数据中心项目不再与公职人员签署保密协议 dev.to NVIDIA 发布四个开放 Cosmos 3 仿真模型 dev.to 真实 MCP 客户端击穿我们服务器的五种方式 theverge.com Airbnb 正在开发 AI 旅行 agent,计划 2027 年上半年推出 wechat 这可能是美国AI最魔幻的一天。
多元动态
登录 注册
行业动态 engadget.com · @staff@engadget.com (Igor Bonifacic)

Suno 与华纳、BMG 合作训练 v6 AI 音乐模型

原文 我的收藏

Suno 发布 v6 系列模型,包括 v6、v6-wild 和 v6-mini,首次使用来自唱片公司的授权数据训练。新模型支持自然语言编辑歌曲片段、生成混音、从音频/图像/视频生成音乐,并增强情感理解。

AI推荐理由 Engadget 报道 Suno 发布 v6 系列模型,首次与华纳和 BMG 合作训练,提供具体功能细节(如自然语言编辑、混音、多模态理解)。来源可信,信息具体,对 AI 音乐应用有参考价值。

70

正文

News AI Suno trained its v6 AI music models with help from Warner and BMG It's the formal start of the company's partnership with some labels. By Igor Bonifacic Sept. 9, 2026 5:00 am EST Igor Bonifacic for Engadget Add Engadget on Google: Preferred Source Google Discover AI music generator Suno has just introduced its new v6 family of models. The models mark the first time the company has worked with labels to train a new set of AI systems. As of today, the v6 series consists of three models: v6, v6-wild and v6-mini. The first of those is Suno's new flagship offering. The company describes v6 as "reliable, precise," and capable of delivering consistently "polished music across every genre and style." For users who want "less predictable" outputs, there's v6-wild. Suno says it's designed for experimentation. "What we've heard time and time again [from musicians and producers] is that they're often not looking for the single best output of the music model," says Jack Brody, Suno's chief product officer. "They're actually looking for something that helps them think more expansively about the kind of music they could make. They're looking for a brainstorming partner, and so v6-wild is our most creative model yet." Finally, there's v6-mini, which is the one model Suno is making available to free accounts; the other two require either a Pro or Premier subscription to access. According to Brody, the new models are faster than the ones they replace, allowing users to listen to the tracks they generate within seconds of writing a prompt. They also offer new features, starting with the ability to edit specific sections of a song using only natural language. Additionally, you can now generate mashups from multiple sources in a single request. So, for example, you can tell the models to pull vocals from one of your generated songs and drums from another, then add new lyrics on top. Sampling is more precise too, allowing users to isolate and generate a new beat through a single text prompt. Separately, the new models offer better multi-modal understanding, allowing them to generate songs from other audio, as well as images and video. At the same time, you can now describe a feeling you want a song to evoke and the models will do their best to create something appropriate to that emotion. You can now also replace a single lyric without generating a song all over again. When I asked Brody to describe how the new models work, he declined to share specifics, saying only that the v6 family incorporates training data Suno had not used before. "The v6 family has been trained from the ground up with a new set of data that differs from the data that went into our previous models," he said. "That includes licensed data from our partners and user data, but most importantly all of the iteration and R&D that has gone into fine-tuning these models in ways to make them subjectively the best models on the market." You might recall that in 2024 a group of record labels made up of the industry's largest players, including Warner Music, sued Suno and Udio, accusing the companies of copyright infringement at a "massive scale." In 2025, Warner dropped its lawsuit against Suno in exchange for a licensing agreement that it said would gave its artists "full control" over how their voices and music are used in AI-generated tracks. Brody said v6 is the formal start of the company's relationship with labels like Warner and BMG. "From the day it launches on September 9, this partnership starts generating revenue for our partners," he said. "[...] it enables new product experiences to be developed with artists and our partners that allow people to create music and remix music from participating artists, which will in turn create additional revenue streams for the artists who participate." Whether those partnerships are enough to legitimize Suno remains to be seen. A 2026 leak found that the company had scraped decades' worth of music and podcasts from YouTube, Deezer and other platforms to train its previous models. Not by accident, Suno is retiring those older models as the v6 family rolls out.

来源报道 · 2 篇 · 以下链接均前往原始媒体

staff@engadget.com (Igor Bonifacic) Suno 与华纳、BMG 合作训练 v6 AI 音乐模型 09-09 · 原文发布
techurls::Engadget Suno 与华纳、BMG 合作训练 v6 AI 音乐模型 09-09 · 原文发布
← 返回精选动态 原文

更多内容