Repository navigation
Sound Effects
Sound effects are an optional layer beside the narration: a creaking door, distant thunder, rain under a scene. They are off at the start of every narration, and nothing plays until you turn them on in the reader's Sounds tab. Everything happens on your computer; no text leaves it.
Open a document and choose the Sounds tab in the side panel (on a phone, open Contents and pick Sounds). The switch at the top turns effects on for this narration only: opening another document, or reaching the end, turns them off again. The volume and the intensity (Minimal, Balanced or Immersive) are remembered on this device. The Audio mixer button in the player has the same switch and volume as a Sound effects channel, beside the narration and background music levels.
Effects never change the narration: they play on a separate audio path with their own volume and a limiter, so a loud effect cannot clip or drown the voice. If a sound fails to load, it is skipped and the narration carries on.
Analyse this chapter looks at the chapter you are viewing and suggests effects where the text describes a sound that is really happening: "the door creaked open", "thunder rumbled", "rain pattered on the roof". Metaphors ("a storm of emotions"), memories, questions, habits and things characters merely say are left alone, and the setting keeps suggestions few and spaced out. A false effect is worse than a missed one, so a quiet chapter often gets none.
Each suggestion shows its sentence, the sound chosen, where in the sentence it falls, and buttons to Preview, Accept, Dismiss or Replace the sound. Nothing plays until you accept it. A dismissed suggestion does not come back when you analyse the chapter again; accepted ones and the effects you added yourself are never touched by a new analysis.
The analysis uses word rules on your computer, not a language model, and the text of a document is only ever data to it: instructions written inside a book cannot make TTSNative do anything.
Add an effect opens the sound library: browse by category, search by a word (rain, door, bell), preview, and choose Use to place it at the sentence being read (or, when the narration isn't on this document, the first sentence of the chapter). An ambience such as rain plays under the next few sentences; a one-shot such as a knock plays once.
You can also select text in the document and choose Sound effect beside it: the sound you pick starts at the first sentence you selected, and an ambience plays across all of them. This works on a phone too.
Each placed effect can be turned off and on, adjusted (at the start or the end of its sentence, with a delay, its own volume, and for ambience how many sentences it lasts), replaced with another sound, or removed. Sentences that carry an effect are tinted in the text, and clicking an effect's sentence in the panel jumps to it.
Effects placed by a reading experience show as From the reading experience. Adjust one and it becomes yours; remove one and it stays gone until you apply the experience again; turning the experience off removes the rest. Effects stay off until you turn them on, as always.
Narration is read one sentence at a time, so an effect lands exactly at the start or end of its sentence. A position inside a sentence is an estimate from where the words fall in the text, usually within half a second. Seeking, pausing, changing speed or voice and skipping chapters all keep effects in step: an effect whose moment has passed is skipped rather than played late, and an ambience stops when you leave its sentences and starts again when you return.
The bundled library has 36 sounds across weather, nature, animals, doors, footsteps, household,
fantasy, impacts, vehicles, atmosphere and machines. Every one was synthesised from a recipe in the
repository, so there are no recordings and nothing to license beyond the project's MIT licence;
apps/web/public/sfx/CREDITS.md lists them. Importing your own sounds is not supported yet.
Eight of them are marked Experimental in the library: the dog, owl, wolf, horse, birdsong, crowd, breaking glass and clashing swords, whose real sources the synthesis cannot yet make convincing. They are never suggested automatically, and a reading experience never places them; you can still add one by hand if it suits your book. Sounds never play louder than the narration: by default they sit well under it, and even at the loudest settings an effect only reaches the narration's own level.
Effects, your choices about suggestions and the analysis record are stored with the document in TTSNative's database on your computer, and deleted with the document. Nothing is written into the original file. If a sentence's text changes (for example after a section correction), its effect is shown with a warning and not played until you place it again.
Exported audio does not include sound effects yet.