Seed Audio 1.0 generates 120-second scene audio from one prompt

• Seed Audio 1.0 is framed as a single-prompt scene-audio generator: dialogue, sound effects, and background music generated together instead of stitched across three tools, according to hasantoxr's launch thread.
• The advertised ceiling is 120 seconds per generation, with reference audio for voice consistency and multiple characters in one prompt, per hasantoxr's feature list.
• The clean workflow is audio first, video second: BytePlus pairs Seed Audio with Seedance so a generated track drives video timing, according to hasantoxr's Seedance handoff.
• Creators are already testing production-shaped loops, from a Seedance guitar-sync workflow by techhalla to magnific's branded audio-ad demo.

A small line in hasantoxr's creator-use post says people are already turning full ebooks into audiobooks, not just generating isolated voice lines. hasantoxr's access note says enterprise API access is open and individual creators can test in the BytePlus console through the linked console. fabianstelzer's post stacks GPT Image 2, Seedance 2, and Seed Audio into a black-and-white cartoon clip.

hasantoxr describes Seed Audio 1.0 as a scene generator: one prompt produces dialogue, sound effects, and background music. The workflow target is obvious: fewer handoffs between voice tools, SFX libraries, music generators, and an editor.

The same thread sits next to hasantoxr's broader game-dev reaction, where he says ideas that used to stay in his head can now become playable in minutes. For filmmakers and animators, Seed Audio points at the same compression: describe the scene track first, then build visuals around it.

The feature list is short enough to quote as structure:

• Up to 120 seconds of full scene audio per generation.
• Reference audio that keeps a character voice consistent across a longer story.
• Multiple characters with distinct voices from one prompt.

hasantoxr also claims creators are already using the model for full ebooks, audiobooks, and multi-character scenes. The phrase to keep is “one voice, many roles,” because it turns reference audio into a casting tool instead of a single-line cloning trick.

BytePlus is pairing Seed Audio with Seedance in an audio-first video workflow:

1. Generate the scene audio.
2. Feed that audio into Seedance as the reference track.
3. Let the video follow the timing.

According to hasantoxr's access note, enterprise API access is open now, while individual creators can test through the BytePlus console link.

techhalla tested Seedance 2.0 on Leonardo with an audio-driven guitar-sync clip. The useful part is the order of operations:

• Start with a short clip using a song segment and a guitarist image as references.
• Set the visual tone first, even if some melody details miss.
• Repeat the process for the rest of the song parts.
• Tweak prompts and ask for extra cuts when needed.
• Assemble the clips in an editor and lock cuts to the beat.
• Export the audio pass, then use an AI music generator to make a custom track for the video.

The prompt screenshot in techhalla's thread names Seedance 2.0, HD 16:9 video, an attached audio reference, and a performance instruction for hand motion, body movement, fast cuts, and no watermark.

fabianstelzer posted a short black-and-white cartoon labeled “Fable's Follies (1928),” made with GPT Image 2, Seedance 2, and Seed Audio.

The stack is the point:

• GPT Image 2 for image generation.
• Seedance 2 for motion.
• Seed Audio for the sound layer.

magnific showed Seed Audio 1.0 inside a ChatGPT plus Magnific MCP workflow for branded audio ads.

The demo flow:

• Give the agent the brief and brand guidelines.
• Ask the Magnific MCP to create an audio clip with Seed Audio 1.0.
• Generate the final branded audio asset from that agent session.