Suno v6 vs Lyria 3.5: Two Ways of Making AI Music Legitimate

If you listen to the marketing, Suno v6 and Lyria 3.5 are the same product: an AI that writes full songs, and now one that does it on licensed music so you are not building on someone else's copyright problem. Google DeepMind's Lyria 3.5 launched on July 29, 2026 inside Flow Music and reached all Gemini users on September 4. Suno v6 arrived five days later, on September 9, trained in partnership with Warner, BMG and Believe. Same claim, but the two models sit at opposite ends of almost every decision that follows: where you run them, how long a song can be, who pays whom, and what happens to the track after it exists. This comparison is about those differences, because "licensed" stopped being a differentiator the moment both of them had it.

The short version, for someone deciding in one paragraph: Suno v6 is a consumer studio with the biggest editing feature set in the category — section rewrites, lyric-level swaps, mashups, audio sampling — sold as a subscription, with label revenue-sharing built into the price. Lyria 3.5 is Google's answer: a model that caps songs at about three minutes, lives inside Flow Music and the Gemini app, comes with SynthID watermarking on every output, and is also the only one of the two you can reach through a developer API today. Neither has an independent listening score yet, so the matchup is decided on access, control, provenance and genre character rather than on any measured quality gap.

The shape of each product

Suno v6 is a subscription consumer product. Pro (about $10 a month, less billed annually) and Premier (about $30 a month, less billed annually) unlock the flagship Suno v6 and the exploratory Suno v6-wild; a free tier runs Suno v6-mini. The models replaced every older Suno model on September 9, and Suno has not opened a public developer API for v6 — the licensing deals run through Suno's own app, and outputs are served and stored inside it.

Lyria 3.5 is spread across Google's surfaces. Its home is Flow Music, Google's song-creation platform (the former Riffusion/ProducerAI product Google acquired in February 2026), which has a free tier plus paid tiers Google prices roughly at $6, $18 and $48 a month. Since September 4 it is also available to Gemini app users worldwide and — the part that matters to developers — through the Gemini API, Google AI Studio and Google Vids. Google has not published per-request pricing specific to the Lyria 3.5 generation, so anyone planning to build on the API should treat the cost as unconfirmed until Google posts a rate card.

• Access — Suno v6 in Suno's apps, subscription only, no developer API vs Lyria 3.5 in Flow Music and the Gemini app, plus the Gemini API and AI Studio.

• Length — Suno v6 length not capped at launch (Custom/Auto duration, Max Mode for long-form consistency) vs Lyria 3.5 songs of 30 seconds to about three minutes.

• Editing — Suno v6 section rewrite, single-lyric swap, mashups, sampling from audio vs Lyria 3.5 Selective Section Painting, lyric rewrite/translation, instrument swaps inside Flow Music.

• Provenance — Suno v6 trained with Warner, BMG and Believe on licensed catalogs, revenue shared vs Lyria 3.5 trained on what Google says is licensed data, SynthID watermark on every track.

• Price — Suno v6 from free (v6-mini), Pro ~$10/mo, Premier ~$30/mo vs Lyria 3.5 free tier in Flow Music, paid tiers ~$6–$48/mo, API pricing unconfirmed.

• Independent scores — none for either model at time of writing; all quality claims are vendor-reported.

Suno's figures above are Suno's published plan pricing and feature descriptions; Lyria 3.5's are Google's published feature set and Flow Music pricing. Neither side has an independent benchmark of the other, and both companies' vocal-quality claims have been questioned in early community tests.

The provenance claim, compared honestly

This is the axis where the two companies' stories sound alike but are not the same weight. Suno v6 has named, contracting partners: Warner settled its copyright lawsuit against Suno in November 2025 and then helped build the model, supplying its own catalogs for training; BMG's music was added; Believe signed its global deal on September 8. A portion of Suno's subscription revenue flows to those three from day one, and they decide how to pay artists. That is a contract, not a press release, and it is why Suno retired its older unlicensed models — the ones Universal and Sony are still suing over, and that a Munich court ruled against on July 31 in the GEMA case.

Google's claim is thinner and harder to verify. Google says Lyria was trained on licensed data, and SynthID watermarking means every output carries provenance whether you want it or not. But Google has not named its licensors, and it is itself defending a lawsuit from musicians who allege Lyria 3 was trained on copyrighted YouTube recordings — a case Google has moved to dismiss. For a creator choosing between the two on "who is safest to use commercially," the honest answer is that Suno v6's safety rests on disclosed label deals with unresolved litigation behind it, while Lyria 3.5's rests on an assertion. Both are meaningfully safer than the 2024 generation of music models; neither is legally bulletproof.

The capture above is OrcaRouter's own earlier explainer of the Lyria 3.5 launch, which covered Google's rollout in detail — the four claimed improvements (musicality, lyrics, vocals, creative control), the Flow Music editing suite, and the September 4 Gemini expansion. This piece leans on that coverage for the Lyria side of the matchup.

Length and control: where they diverge most

Lyria 3.5 is deliberately a short-form model. Google gives you granular control over tempo and duration in the roughly 30-second-to-three-minute band, plus genre, BPM, key/scale, mood and song-structure settings, and image-to-music from up to ten reference images. Editing is done by "painting" a section of the waveform and regenerating just that part. That makes Lyria 3.5 strong for a specific job — a cue, a bed, a verse-and-chorus piece that needs to sit at a given tempo — and its character is widely described as cinematic and orchestral, with pop vocals as the weaker suit.

Suno v6 runs the other direction. It does not publish a hard length cap and pushes toward complete, long-form, vocal-led songs, with a Max Mode toggle aimed at keeping consistency across extended tracks, covers and voices. Its editing model is textual rather than spectral: you tell the song what to change in words ("make the chorus a gospel choir," "swap this one lyric," "add the drums from this other track"), and it rewrites only what you pointed at. Suno's genre strength has always been full pop and rock songs with expressive vocals, and v6 keeps the personas, voices and covers machinery that Lyria has no equivalent for.

If your deliverable is a two-and-a-half-minute song with a proper arc, both get you there. If your deliverable is a ten-minute piece or a vocal performance you can steer line by line, only one of them is built for that.

What each costs, and where the money goes

The price tags look close — both have a free tier and paid tiers in the $6–$48 band — but the money flows differently. A Suno v6 subscription buys you generation plus label-licensed training data plus, on paid tiers, assigned rights in the outputs; Suno passes a share to Warner, BMG and Believe, and the labels pass it to artists. A Flow Music or Gemini subscription buys you generation plus SynthID provenance, with Google's licensing costs absorbed into the subscription rather than disclosed. For an individual creator the difference barely matters. For a company that needs to show a rights chain before it will let AI music near a commercial product, the Suno v6 structure gives you a named counterparty to point at, and that is not nothing.

Which one to open

Pick Lyria 3.5 when you already live in Google's ecosystem, when you want instrumentals and cinematic material with tempo-and-key control, when a three-minute piece is the right size for the job, or when you are a developer who needs an API — Lyria 3.5 is the only model in this matchup you can call today, via the Gemini API.

Pick Suno v6 when you want complete, vocal-led songs, when you need long-form or want to iterate by rewriting sections and single lyrics rather than regenerating, or when a disclosed label licensing deal matters to your commercial use. The free Suno v6-mini is also the cheapest way to A/B the two camps: generate on v6-mini, generate the same prompt in Flow Music's free tier, and let your own ears — not either vendor's marketing — make the call.

Two open questions will decide this matchup over the next few months. First, Google has not priced Lyria 3.5 on the Gemini API, and until it does, the "developer-accessible licensed music model" crown is conditional. Second, Suno v6 has no API at all, so the only way to reach it is a subscription. The moment both have published API prices, this stops being a product comparison and becomes a per-generation cost comparison — and the way to run that comparison is the same playbook any team already uses for model upgrades: put the challenger and the incumbent on parallel test paths, watch quality and cost per generation, and let real output decide. It is the playbook a pass-through router like OrcaRouter industrializes for the 200+ API models behind one endpoint, and it applies here whether or not music models ever sit on that endpoint. Until then, the two are answering different briefs, and the right one depends on whether you are writing songs for an audience or scoring something for a client.