How Many Voice Lines Does an Indie Game Need?

August 6, 2026

Answer first. Indie games that ship voice acting tend to land in one of three brackets: a key-scene build voices roughly 30 to 80 lines, a partial-voice build voices roughly 150 to 400, and a full-voice narrative game starts around 1,500 and climbs. Treat those as planning brackets for your own scoping, not an industry survey.

The bracket matters more than any number inside it, because it decides how much of your game the voice has to carry. Below we walk the count in the order that works, then convert it into words, which is what AudioProducer meters. For the production workflow itself, how to voice NPCs in your indie game covers casting and generation end to end.

Count scenes before you count lines

A line count you arrive at by scrolling your script is a guess, and it is usually low. Scenes stay stable early in development while the dialogue is still moving around, so count those.

Open your outline and count the moments where a character speaks to the player memorably. Give each a rough line count: a short exchange is 6 to 10 lines, a real conversation is 20 to 40, a boss confrontation with interruptions can reach 60. Multiply and the number is anchored to structure instead of to whatever is written today.

It is also where you find the dialogue you never think about: shopkeeper haggling, the tutorial companion, the chatter in the hub town. None of that appears in an outline, and together it is often the largest block of lines you ship.

The three tiers most indie games ship

Key-scene voice, 30 to 80 lines. The opening, one midpoint turn, the ending, and maybe a recurring character who greets the player. Everything else stays text. This is the tier where a solo developer finishes, and we suggest it for a first project.

Partial voice, 150 to 400 lines. Every named character speaks in their first scene, and quest turn-ins and emotional beats are voiced. Repeat interactions fall back to text. Most narrative indie games that feel voiced are sitting here.

Full voice, 1,500 lines and up. Every written line has audio. The count is not the hard part here, the maintenance is: every script revision means regenerating lines, renaming files, and re-testing triggers.

Heavy branching pushes these numbers hard, because a branch that reads as one scene to the player is several scenes to you. Voices in a Twine game and voice acting in a visual novel both count branched dialogue, and the arithmetic transfers to any engine.

Which lines earn a voice, and which never do

Lines worth voicing share one property: the player hears them once, or rarely, at a moment that carries weight. First meetings, quest turn-ins, death and failure states, and the single scene your game is actually about.

Lines that do not earn a voice are the ones the player triggers over and over: menu confirmations, item descriptions, the tutorial prompt that fires on every crafting screen. A line heard forty times stops being performance and becomes noise, so voicing it costs you quota and patience at once.

Lore entries and codex pages sit in between. They read as narration, and one narrator voice handles them well, the same way a tabletop sourcebook does.

Barks are where the count runs away from you

Barks are the short repeated lines that fire during play: a guard noticing you, a hit reaction, a shopkeeper greeting. On your outline they look like six lines. In the build they are forty, because each one needs several readings or the player hears the loop inside a minute.

Here is the honest constraint. AudioProducer renders the same text the same way, so a second reading does not come from adjusting a setting. There is no pitch control, no speed control, and no randomize button. A variant is a different line of writing, so five ways for a guard to notice the player means five written lines. Per-line dialogue emotion is the one per-line lever we do have, and it changes delivery rather than wording, so it carries a repeated line only so far.

Multiply before you budget. Twelve bark types at five variants is sixty lines, and those sixty are invisible on any outline. The bark workflow itself is covered in the NPC voicing guide.

What your line count costs in words

AudioProducer meters words, so convert before you plan a budget. Game lines are short: most sit between 8 and 20 words, and 14 is a reasonable planning midpoint.

A worked estimate for a three-hour narrative game at the partial-voice tier: 18 story scenes averaging 14 lines gives 252 lines, plus 12 bark types at 5 variants gives 60, for 312 lines total. At 14 words each that is roughly 4,400 words.

Against that, the free tier is 1,200 words with no card required, covering about 85 short lines. That is a vertical slice, one scene plus its barks, and it is the right thing to generate first. Paid plans start at $39.99 per month.

One number people miss: regeneration is metered again. If you generate 4,400 words and then decide the merchant sounds wrong, recasting costs those lines their words a second time. Settle your cast on a sample scene first.

What 300 lines means at export time

AudioProducer's export unit is the chapter. One button renders a whole chapter into a finished MP3, and each chapter downloads as its own file. There is no per-line export. A game needs one file per line, so the working approach is to make each voice line, or each tight group that always plays together, its own chapter and download chapter by chapter. A 312-line script is roughly 312 chapters. Better to hear that now than at export.

Two things reduce the pain: group aggressively, and build in passes, one character at a time, so you export 30 chapters at a go.

Past the download, the pipeline is yours. There is no Unity or Godot plugin, no API for voice lines, and AudioProducer does not distribute or host anything. You get MP3s you own and wire them in yourself. Import works from EPUB or pasted text, so a script living in a spreadsheet gets pasted in.

The short version: count scenes, pick a tier, multiply your barks by their variants, then convert at roughly 14 words per line. Generate one scene and listen to it before you scale. Projects rarely stall on the line count. They stall on meeting the export shape after the script is written. For adjacent ground, see interactive audio stories and GameLit audiobooks with AI.

Frequently asked questions

How many voice lines does a short indie game need?
A key-scene build covers most short games at roughly 30 to 80 lines: the opening, one midpoint turn, the ending, and perhaps a recurring character who greets the player. Everything else stays text, and players rarely miss it.
Can I work out my budget on the free tier?
Yes. The free tier is 1,200 words with no card required, which is about 85 short game lines at a 14-word planning midpoint. That covers one scene plus its barks, which is the right first thing to generate and listen to. Paid plans start at $39.99 per month.
Does AudioProducer export one audio file per line?
No. The export unit is the chapter. One button renders a chapter into a finished MP3 and each chapter downloads as its own file, so for a game you make each voice line, or each tight group that always plays together, its own chapter and download chapter by chapter. There is no per-line export, no Unity or Godot plugin, and no API for voice lines, so wiring the MP3s into your engine is your own pipeline.

Related posts