Skip to content
BytePlusBytePlusBytePlus ModelArk for ComfyUI

Seed Speech TTS and ASR

Template
Seed Speech TTS and ASR, in the ComfyUI template browser under BytePlus
Workflow file
Seed Speech TTS and ASR.json
BytePlus nodes
Seed Speech TTS, Seed Speech ASR
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 spoken sentence, its transcript and its subtitles. The template reads “Welcome to BytePlus Seed Speech. This sentence was synthesized, then transcribed again by the ASR node.” in the TTS 2.0 voice Stokie, then sends that audio straight into the ASR node. The transcript that comes back is the original sentence, which shows both nodes working together.

The workflow

Section titled “The workflow”
The Seed Speech template: TTS output going to Preview Audio and into the ASR node, whose text and srt outputs go to two Preview as Text nodes
Five nodes: Seed Speech TTS, Preview Audio, Seed Speech ASR and two Preview as Text.
  1. Seed Speech TTS uses model seed-tts-2.0 and the voice en_female_stokie_uranus_bigtts to speak the text. context_text is “Speak in a bright, friendly tone.”
  2. Preview Audio plays the speech.
  3. Seed Speech ASR takes the same audio, uses the fast model seed-asr-fast with language on auto, and returns the transcript, the utterance timings and an SRT.
  4. Preview as Text (two nodes) show the text output and 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 Speech TTS and ASR under BytePlus.
  3. Press Run. The audio, the transcript and the subtitles appear in the preview nodes.

Settings worth changing

Section titled “Settings worth changing”
  • text (TTS): write your own sentence. Pick a voice that supports its language.
  • voice: choose another TTS 2.0 voice. To use a cloned voice instead, switch model to seed-icl-2.0 and fill custom_speaker_id (see Seed Voice Clone).
  • context_text and emotion: steer the delivery with a style instruction or an emotion such as happy or sad. context_text is not billed.
  • hotwords (ASR): list names or terms to favour, one per line. The template has “BytePlus” and “Seed Speech”. Hotwords work only with the Chinese-English model, language auto or zh-CN.
  • model (ASR): seed-asr-2.0 covers more languages and runs as submit and poll. It uploads connected audio to Comfy.org storage first, so log in to your Comfy.org account in ComfyUI. The older seed-asr-1.0 stays in the node, but no template uses it.

Example result

Section titled “Example result”
The graph after the run with a 6.4 second waveform, the transcript and two subtitle entries
Example result: the transcript matches the spoken sentence.
The speech from this example.

Time and cost

Section titled “Time and cost”

Each node sends one request to Seed Speech. The TTS seed is not sent to the API; it only decides whether ComfyUI runs the node again. Its control is set to randomize, so every run speaks the sentence again and bills again. Set it to fixed and ComfyUI reuses the cached result of an unchanged workflow without billing.

Section titled “Related”