📐 Visual Tag Builder

Designing a voice from a text description is powerful, but writing that description freehand is fiddly - you forget which attributes the model understands, or you word "elderly man, low pitch, British" in an order the model doesn't like. This node turns that into point-and-click. You pick gender, age, pitch, accent, and a few more from dropdowns, and it assembles a properly-ordered instruction string for you. It was built for OmniVoice's voice-design workflow but is generalized to feed any tag-based prompt.

Its output is just a string. That string becomes the voice-design instruction for an OmniVoice engine (or gets used wherever a structured tag/attribute prompt is handy). So this node is an authoring convenience - it doesn't generate audio, it generates the description that a voice-design engine then acts on.

How it works

Each dropdown maps to a voice attribute, and the node stitches your selections into one instruction with a consistent, preset-driven order. Leave any field on None and it's simply omitted, so you can be as specific or as loose as you like - "female, elderly, low pitch" or just "male, whisper." The preset-driven ordering matters more than it sounds: models respond better to attributes in a stable order, and the builder handles that so you don't have to think about phrasing.

The inputs and outputs that matter

All the fields are dropdowns you set to taste; the ones you'll actually use most:

• gender - None, male, female.
• age - child, teenager, young adult, middle-aged, elderly.
• pitch - very low through very high.
• accent - a menu of accents (american, british, australian, indian, and more).

Rounding it out: style (currently whisper as the non-None option), dialect (a set of Chinese regional dialects), output_language (English or Chinese - which language the instruction is written in), and instruct_text, a free-text field to append anything the dropdowns don't cover.

The single output is string - the assembled instruction. Wire it into the voice-design instruction of an OmniVoice engine, or into the 🎨 Voice Designer node's instruction, or anywhere a descriptive prompt string is useful.

Installing it

Bundled with the pack. ComfyUI Manager → search "TTS Audio Suite" → install → restart. Manual:

cd ComfyUI/custom_nodes
git clone https://github.com/diodiogod/TTS-Audio-Suite.git
cd TTS-Audio-Suite
python install.py

No model download - it's a pure text-assembly node, so it loads instantly. The OmniVoice model it typically feeds downloads separately when you actually generate.

Common issues & troubleshooting

The voice didn't match the description. This node only builds the instruction - whether the voice matches is the engine's job. Make sure the string is actually wired into a voice-design-capable target (OmniVoice in voice-design mode, or a Voice Designer setup), and that you're not accidentally feeding it into the spoken-text input.

The dialect options are all Chinese. That's expected - the dialect dropdown targets Chinese regional dialects specifically. For accents in other languages, use the accent field instead.

I need attributes that aren't in the dropdowns. Use instruct_text to append free-form description. The builder covers the common structured attributes; anything else goes in that field. And because the node supports reusable custom presets, if you keep re-typing the same extra attributes it's worth saving them as a preset.

Nothing appears in the output. If every dropdown is None and instruct_text is empty, there's nothing to assemble. Set at least one attribute.