SeedAudio 1.5

What is SeedAudio 1.5?

SeedAudio 1.5 (also written "seed audio 1.5") is an announced audio generation model from ByteDance Seed, presented on this site as its "next audio model for complete scenes" and as an AI audio generator. It is designed to produce dialogue, music, ambience, and sound effects in one pass, and to return those elements as separate stem tracks. The model is not yet available: the site marks it "Coming Soon" in the model menu, and the generator that runs on the site today is Seed Audio 1.0. According to the published specification shown on the site, SeedAudio 1.5 raises maximum output length from two minutes to six minutes and the number of supported audio references from three to six.

How SeedAudio 1.5 works

Generation is prompt-first, with optional controls and templates. Four input modes are published: text (T2A), text plus audio (TA2A), text plus video (TV2A), and text plus audio plus video (TAV2A). The site states that text-plus-video and text-plus-audio-plus-video modes support dubbing and sound design around existing footage, and that the model uses reference-guided voices and timestamp control. The workspace prompt field accepts up to 2,048 characters, alongside a prompt optimizer and ready-to-use scene formulas organized by format. Each formula is a written scene description covering character voices, environmental beds, foley actions, and musical cues; clicking a card loads it into the workspace.

Main features

• Layered output: dialogue, ambience, SFX/foley, and music delivered as four stem tracks.
• Length: up to six minutes per generation once 1.5 launches, up from two minutes in Seed Audio 1.0.
• References: up to six audio references, up from three.
• Languages: 30 languages plus regional variants.
• Input modes: four published modes spanning text, audio, and video.
• Voice and dubbing: reference-guided voices and video-aware dubbing.

Pricing

The pricing page lists two paid tiers plus a free option, with a monthly/yearly toggle and yearly billed as "2 months free."

• Free — $0: 10 free credits, up to about 8 seconds per generation, max 2 minutes of audio per generation, API access, priority support. Daily check-in credits are also offered.
• Pro — $19.99/month, or $16.66/month billed yearly (listed saving $39.98): 30,000 credits per year, up to about 400 total audio minutes, everything in Free, reference voice uploads, priority support, max 2 minutes per generation, and early-bird SeedAudio 1.5 access on shared credits.
• Max — $49.99/month, or $41.66/month billed yearly (listed saving $99.98): 90,000 credits per year, up to about 1,200 total audio minutes, everything in Pro, a larger annual production buffer, and team/agency-friendly usage.

The page states that plan prices are unchanged, that credits work today on the Seed Audio 1.0 generator, and that one shared balance covers both the web app and API. A limited-time early-bird offer is displayed with a countdown; the site says orders placed now lock in early-bird pricing.

Common use cases

Per the site, SeedAudio 1.5 is aimed at video dubbing, localized ads, game scenes, podcasts, and longer narration where dialogue, ambience, effects, and music need to align to a scene. The sample scene formulas cover stadium sports commentary, a rainy-apartment radio drama, a travel podcast, and a Mandarin sports-car commercial.

Limitations

• SeedAudio 1.5 is not yet available; the generator on this site still runs Seed Audio 1.0, and the six-minute, six-reference, and four-input-mode figures describe the announced model.
• Current limits differ: free generations stay at about 8 seconds, and paid Seed Audio 1.0 generations stay at 2 minutes.
• On the current Seed Audio 1.0 generator, most failures come from input length or content policy: reference audio longer than 30 seconds, prompts with sexual, explicit, violent, or graphic content, or prompts naming specific real people or protected characters.
• Commercial use depends on the terms of the access channel used to generate the audio; license, content restrictions, and output rights should be checked before use in ads, games, podcasts, or paid media.
• Seed Audio 1.0 accepts text, up to three audio references, or one image — not the four input modes described for 1.5.

Summary

SeedAudio 1.5 is a coming-soon ByteDance Seed audio model for scene-level generation, combining dialogue, ambience, foley, and music in one pass with stem separation, four input modes, up to six minutes of output, six audio references, and 30 languages. Until it launches, this site generates with Seed Audio 1.0 and sells early-bird credits through Free, Pro ($19.99/month, or $16.66/month yearly), and Max ($49.99/month, or $41.66/month yearly) plans on a shared web-and-API balance.