Skip to content
BytePlusBytePlusBytePlus ModelArk for ComfyUI

Podcast Clip

Template
Podcast Clip, in the ComfyUI template browser under BytePlus
Workflow file
Podcast Clip.json
BytePlus nodes
BytePlus LLM, Seed Speech TTS ×2, Seed Speech ASR, Seedream 4.5 & 5.0
Keys needed
ModelArk API key, Seed Speech API key
Also uses
Concatenate Audio, Extract Text, MarkdownNote, Preview as Text, Save Audio (Advanced), Save Image, Text (Multiline) (built into ComfyUI)

What it makes

Section titled “What it makes”

A short audio episode with a host and a guest, an SRT subtitle file for it and a square cover picture. It is for creators who want a quick explainer clip on any topic. The idea: one language model writes the dialogue and the cover prompt, two TTS voices speak the two roles, speech recognition turns the finished audio back into timed subtitles and Seedream paints the cover.

The workflow

Section titled “The workflow”
The Podcast Clip template in five groups: topic, dialogue, voices, subtitles, cover art
Five groups. Both voices feed the Join node, and its audio goes to Save and to the ASR node.
  1. A Topic text node holds the subject (“Why do cats purr? A friendly two-person explainer”).
  2. BytePlus LLM (Seed 2.1 Turbo) replies with HOST:, GUEST: and COVER: lines. Three Extract Text nodes split them, and Preview as Text nodes show the host and guest lines.
  3. Two Seed Speech TTS nodes, Host voice (an English female voice) and Guest voice (an English male voice), speak the lines with seed-tts-2.0.
  4. Concatenate Audio (titled “Join the two voices”) joins the two clips, guest after host, and Save Audio (Advanced) writes an MP3 to BytePlus/Audio/podcast.
  5. Seed Speech ASR (seed-asr-fast) transcribes the joined audio. Its srt output is shown in a Preview as Text node.
  6. Seedream 4.5 & 5.0 (Seedream 5.0 Flash, 1:1 at 1K) paints the cover from the COVER: prompt, and Save Image keeps it.

Run it

Section titled “Run it”
  1. Add your ModelArk API key and your Seed Speech API key in Settings → BytePlus if you have not yet (API keys). The template needs both.
  2. Open the template browser and choose Podcast Clip under BytePlus.
  3. Type your own subject in Topic.
  4. Press Run. The dialogue appears first, then the audio, the subtitles and the cover.

Settings worth changing

Section titled “Settings worth changing”
  • Topic: add the tone or the audience, for example “for ten-year-olds”.
  • Host voice and Guest voice: pick two clearly different voices. Voice names start with a language code, so keep both in the language of the dialogue.
  • System prompt of the LLM: change the number of sentences, or the language of the dialogue (then pick matching voices).
  • model in Seed Speech ASR: seed-asr-fast handles audio up to 2 hours in one request. The 2.0 and 1.0 models upload the audio to Comfy.org storage and need a login.
  • Cover model and size_preset: switch to Seedream 5.0 Pro, or pick another ratio for the cover.

Example result

Section titled “Example result”
The saved episode as a waveform, the SRT subtitles, the cover art of a sleeping orange cat in a knitted blanket and the written host and guest lines
Example result: an 81 second episode and a cover of a sleeping orange cat.
The saved episode from this example.

Time and cost

Section titled “Time and cost”

The LLM and Seedream nodes bill on your ModelArk account, and the two TTS nodes and the ASR node bill on your Seed Speech account. The seed controls are set to randomize, so every run writes a new dialogue and cover and bills again. Set a control to fixed and ComfyUI reuses the cached result of an unchanged node without billing. Review the subtitles before publishing, especially names and empty entries.

Section titled “Related”