Audio Dialogues

Valheim writes its dialogue on screen and never says it. This mod speaks it: the ravens' tutorials, Haldor's small talk, the dvergr, the runestones, the dreams you wake up from.

This package holds no audio. It is the player and nothing else, 26 KB of it. The spoken lines come from a voice pack installed next to it, Valheim Voices being the first one. Install both, or this mod stays silent and says so in the log.

What gets spoken

Where the line comes from	Who speaks
NPC speech bubbles	Haldor, Hildir, the bog witch, dvergr, fallen valkyries, the shadow people
Ravens	Hugin and Munin, tutorials and guide points
Runestones and tablets	the narrator, including the opening text
Dreams	the voice you hear on waking
Offering stones and tamed companions	the few lines that pass through the message bar

An NPC's voice comes out of the speech bubble, not their feet, and fades out exactly where the bubble stops being drawn. Two NPCs talking at once never cut each other off.

Settings

BepInEx/config/Tchernobill.AudioDialogues.cfg, written on first launch.

Key	Default	What it does
Enabled	true	master switch, no patch is applied when false
Volume	1	on top of the game's sound effects slider
Reverb	1	how much a voice is sent to the game's reverb, as a multiple of what the game sends for its own sounds; 0 keeps it dry
VoicesFolder	empty	absolute path to one voice folder, overriding every installed pack
Language	empty	load only packs declaring this language; empty loads them all
VoiceMessages	true	also speak the lines going through the message bar
DreamFadeIn	1.5	seconds a dream line takes to fade in
CacheSize	64	decoded clips kept in memory
PreloadPrefixes	tutorial_,npc_haldor_greet,npc_hildir_greet	loaded at startup when a pack brings no list of its own
LogTokens	false	log every localization key seen, spoken or not

LogTokens is how you find a line that has no voice yet, and how a game update that renames keys shows up. Each key is logged once per session.

Client side only. Nothing to install on a dedicated server, nothing sent over the network.

Writing a voice pack

A voice pack is a folder, not code. Put a folder named voices anywhere under BepInEx/plugins, up to four levels deep, holding index.json and its ogg files. There is no assembly to reference and no method to call: the index format is the whole contract.

{
  "version": 1,
  "language": "English",
  "preload": ["npc_haldor_random_greet1", "tutorial_bed_text"],
  "clips": [
    { "key": "npc_haldor_random1", "file": "npc_haldor_random1.ogg",
      "duration": 3.42, "hash": "9f2c1ab4", "voice": "haldor", "seed": 1108 }
  ]
}

• key is the localization key without its dollar sign, lowercase.
• file is relative to the voices folder; left out, key + ".ogg" is used.
• duration, hash, voice and seed are informative. Unknown fields are ignored, so the format can grow without this mod being rebuilt.
• preload is optional; without it the PreloadPrefixes setting decides.

Ogg Vorbis, mono. Unity only spatializes mono, a stereo clip would play flat and ignore where the speaker stands. Without an index.json, the ogg files are indexed by their own names, which is enough to try a handful of clips by hand.

Several packs can be installed at once. When two of them carry the same key, the one loaded last wins and the log says so, packs being loaded in path order. That is what lets a pack of a few replacement lines sit on top of a full bank without replacing it.