ComfyUI lab 3: Create clean narration and reusable voices
Treat speech as structured data: text, voice, pacing, pronunciation, sample rate, and provenance.
Listen to this article
On-device voiceUses the voice built into your browser; no article text leaves this page.
Audio narration is not supported by this browser.
Speech is not just an audio file. It is a relationship between text normalization, voice identity, prosody, pronunciation, timing, sample rate, and permission to use the voice. A good workflow makes those decisions visible.
Prepare text for ears
Write what should be spoken, not what looks elegant on a page. Expand ambiguous abbreviations, split long sentences, and use punctuation to signal phrasing. Keep a pronunciation dictionary beside the script for names, products, and technical terms. Generate short paragraphs separately so one mistake does not invalidate five minutes of otherwise good narration.
The graph returns a normal ComfyUI AUDIO value. That matters: it can feed a SaveAudio node, a talking-character node, a waveform analyzer, or a future mixer without being downloaded and uploaded between every stage.
Voice provenance is a required input
Store the voice label, engine/model, reference-clip origin, consent scope, language, and generation date with the output. Never treat “the model allowed it” as consent. For a cloned voice, use a reference you have permission to process and make generated speech clear to the people who will hear it.
Finally measure, do not guess: listen for clipped consonants, breath artifacts, incorrect stress, silence at segment boundaries, and loudness mismatch. Fix the text or voice settings before adding mastering effects.
Run the lab
- Open this graph in Comfy Studio
- Download the exact API-format workflow
- CLI:
app comfy download appnz-hosted-tts
Record the graph JSON, input assets, seed where relevant, model or service version, custom-node commit, and final artifact together. That bundle—not a screenshot—is the reproducible creative unit.
Open-source map
- Open app.nz TTS node — MIT AUDIO-producing ComfyUI node
- F5-TTS — Open-source flow-matching TTS project
- ComfyUI-F5-TTS — Community ComfyUI integration
Creative Systems Labs: 3 of 12. Previous: Build loopable music beds Next: Build a talking character pipeline