Skip to main content
Subtitles · Speech-to-Text

Subtitles · Speech-to-Text · screenshot coming soon

​
What this node does

Subtitles · Speech-to-Text runs a speech-to-text workflow over an upstream video or audio and returns the transcription as COMFYTV_TEXT (subtitle cues). Pick a workflow, wire a clip, press ▶ Run; the transcription runs as a backend workflow (a real model pass on the server), so it needs the corresponding model installed. You must provide either an audio or a video input. If audio is wired it is transcribed directly; otherwise the video is used. Running with neither raises an error.

​
When to use it

  • Generate captions for a talking-head clip, then burn them with Subtitles.
  • Get a rough transcript of a voiceover or interview for editing.
  • Kick off a caption workflow without hand-typing an SRT.

​
Parameters

​
workflow

Which speech-to-text workflow to run. The dropdown lists the installed speech-to-text workflows (labels_for('speech-to-text')), with a sensible default. Each option maps to a backend workflow JSON and its required model.

​
Inputs and outputs

​
Tips

  • Wire subtitles → the subs_text input of the Subtitles node to go straight from speech to burned-in captions.
  • Provide audio when you have a clean audio track — it skips using the video’s soundtrack.
  • The available workflows and their models depend on your install; if the dropdown is empty or Run fails, check the speech-to-text workflow/model is present.

​
Types and bridges

COMFYTV_TEXT, COMFYTV_VIDEO and COMFYTV_AUDIO are ComfyTV project snapshots, not native ComfyUI types. To interoperate with native ComfyUI nodes, insert a Bridge (ComfyTV/Bridge). See bridges.md.
  • Subtitles — burns the generated cues onto the video.
  • Title — a single styled title card.