IndexTTS Generate Simple

IndexTTS Generate Simple is the uncluttered version of the pack's main synthesis node. It does exactly what its big sibling IndexTTS Generate does - same voice cloning, same multi-voice and pause support, same subtitle outputs - but it hides the entire sampling rack. You get indextts_model, a text box, the three reference inputs, unload_model, and seed. That's it.

This is the node you reach for when you just want narration, or when you're dropping TTS into a workflow and don't want 16 widgets demanding your attention. Beginners, start here. If you find yourself wishing for a parameter, the full node is one node to the left and uses these exact same defaults.

What's actually there

• indextts_model - the handle from IndexTTS Model Loader.
• text - multiline script, with the same line syntax as the full node: -2s- for pauses, [voice_name] text for speaker assignment, (start, end) text for timed lines.
• reference_audio / reference_audios / emotions - the voice source. Same priority order (emotions win, then the audio list, then single audio).
• unload_model - free VRAM after the render.
• seed - set for reproducibility.

Under the hood it literally calls the full node's execute with the author's defaults: do_sample on, temperature 0.8, top_p 0.9, top_k 30, num_beams 3, repetition_penalty 10, and so on. So you're not getting a different, cheaper engine - you're getting the exact same render with the dials pre-set.

Outputs and caveats

Same four outputs as the full node: audio, seed, subtitle (JSON), and subtitle_text (readable). Same caveats too - this is still IndexTTS-2 under the hood, so budget ~12GB VRAM, remember it's an English/Chinese model, and avoid apostrophes in your script ("dont", "its"). The one difference that matters: there's no speech_speed control here. If you need to speed up or slow down a take, you'll either grab the full node or time-stretch the audio downstream.

Short, sweet, and honestly the right default for most people.