Skip to content
BytePlusBytePlusBytePlus ModelArk for ComfyUI

Seed Audio

Template
Seed Audio, in the ComfyUI template browser under BytePlus
Workflow file
Seed Audio.json
BytePlus nodes
Seed Audio 1.0
Keys needed
Seed Speech API key
Also uses
Preview Audio, Preview as Text (built into ComfyUI)

What it makes

Section titled “What it makes”

One audio clip of about 16 seconds and its subtitles. The prompt describes a late-night radio host who welcomes listeners over soft rain and a distant jazz piano, and quotes the two lines to speak. Seed Audio 1.0 reads the lines in the preset voice Stokie (Female, American English) and, because subtitles are on, also returns an SRT text with a time range for each sentence.

The workflow

Section titled “The workflow”
The Seed Audio template: one Seed Audio 1.0 node feeding Preview Audio and Preview as Text
Three nodes: Seed Audio 1.0, Preview Audio and Preview as Text.
  1. Seed Audio 1.0 receives the prompt, uses reference_mode = preset voice with the voice Stokie, and creates the audio. enable_subtitle is on in this template.
  2. Preview Audio plays the AUDIO output.
  3. Preview as Text shows the srt output.

Run it

Section titled “Run it”
  1. Add your Seed Speech API key in Settings → BytePlus if you have not yet (API keys). This is a separate key from the ModelArk key.
  2. Open the template browser and choose Seed Audio under BytePlus.
  3. Press Run. The clip appears in Preview Audio and the subtitles in Preview as Text.

Settings worth changing

Section titled “Settings worth changing”
  • text_prompt: describe the voice, the mood, the sounds around it and the lines to speak. Put the spoken lines in quotes.
  • reference_mode: switch from preset voice to text only, audio reference (clone up to three voices from clips, tagged @Audio1 to @Audio3 in the prompt) or image reference.
  • preset_voice: pick another built-in voice while reference_mode is preset voice.
  • speech_rate, loudness_rate and pitch_rate: 0 is normal. For speed and loudness, 100 means twice as fast or loud and -50 means half. Pitch moves by semitones from -12 to 12.
  • generation_count: an advanced input. It runs several takes in parallel, each billed as its own request, and every output becomes a list.

Example result

Section titled “Example result”
The Seed Audio graph after the run with the audio waveform and two subtitle entries below
Example result: a 16.1 second clip and two subtitle entries.
The clip from this example.

Time and cost

Section titled “Time and cost”

The node sends one request per generation and bills it. The seed is not sent to the API and does not make results repeatable; it only decides whether ComfyUI runs the node again. Its control is set to randomize, so every run makes a new take and bills again. Set it to fixed and ComfyUI reuses the cached result of an unchanged workflow without billing.

Section titled “Related”