> ## Documentation Index > Fetch the complete documentation index at: https://docs.decktalk.app/llms.txt > Use this file to discover all available pages before exploring further. # Sound > The DeckTalk mix, with the narration, clip audio, the underscore that ducks under speech, ambience beds, sound effects on cues, underscore markers, and generating sound with ElevenLabs. The video always carries the narration and the audio of any clip section. Everything else is optional and stays off until `[mix]` in `decktalk.toml` names a file. You can bring your own files, or generate them with `decktalk soundscape`. ## The parts of the mix | Part | Source | Behaviour | | ---------- | --------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------- | | Narration | `build/audio/narration.mp3` | Always present. It begins at the start of the first page section and runs continuously. | | Clip audio | each clip file | Plays for the clip's span, trimmed to its picture and faded over twenty milliseconds at both ends. | | Underscore | `mix.underscore` | Loops for the length of the video, fades in and out, and ducks by `underscore_duck_db` under every spoken span. | | Ambience | `mix.ambience` | Plays under sections that set `ambience = true`, with a ramp at each edge and a short pad on either side. | | Effects | `[[mix.sfx]]` | One-shot files placed where a cue resolved, plus an offset. A cue that did not resolve, or a file that is missing, skips its effect with a warning. | Ducking is per spoken section span. Each page section's span runs from its start in the narration to the end of its last word, and each clip's span is its whole length, so the underscore comes back up during the silence at a section's end and stays down under a clip. The duck ramps over `duck_ramp_seconds`, and the ramps and pads are tuning keys under `[audio]` in [Configuration](/reference/configuration#audio). After the parts are summed, the mix is normalized to the targets in `[mix.loudnorm]`. [How it works](/concepts/how-it-works#loudness) describes the pass. The `--no-loudnorm` flag skips it, and `--nomix` produces narration and clip audio only. ## Markers A markers file swells or mutes the underscore at spoken phrases, which gives a music bed some structure without editing it. The scaffold's `media/markers.json` swells the bed at the open. ```json media/markers.json theme={null} { "boost_db": 3, "boost_seconds": 2, "markers": [ { "name": "open", "section": 1, "on": "$start" }, { "name": "answer", "section": 3, "on": "Only two x", "mute_seconds": 0.6 } ] } ``` | Key | Where | Meaning | | ---------------------------------------- | --------- | ----------------------------------------------------------------------------------- | | `boost_db` | top level | How far each marker swells the bed. The default is 3 dB. | | `boost_seconds` | top level | How long the swell lasts. The default is 2 seconds. | | `name` | marker | A label for the report. | | `section` | marker | The section number the phrase is in. | | `on` | marker | A spoken phrase, `$start`, or `$end`, matched as a cue is. The default is `$start`. | | `occurrence`, `case_sensitive`, `offset` | marker | Refine the match as they do for a cue. | | `mute_seconds` | marker | Silence the bed for this long first, and swell after. | | `_comment` | top level | Ignored. | A marker whose phrase cannot be found is skipped with a warning. The mute and boost ramps are `marker_mute_ramp_seconds` and `marker_boost_ramp_seconds` under `[audio]`. ## Generating sound The `soundscape` command turns the prompts in `[soundscape]` into files with the ElevenLabs sound and music endpoints. Ambience and music write to the paths named in `[mix]`, or to `build/sfx/ambience.mp3` and `build/music/underscore.mp3` when `[mix]` names none, and each effect writes to `build/sfx/.mp3`. ```toml decktalk.toml theme={null} [soundscape.music] prompt = "calm, minimal instrumental underscore for a lesson: soft piano and warm pads, slow, no drums" seconds = 120 [soundscape.ambience] text = "quiet lecture hall room tone, distant HVAC, no voices, steady" [soundscape.sfx.tick] text = "soft interface click, single, short, no reverb" ``` ```console theme={null} decktalk soundscape --dry-run # prints every request without sending it decktalk soundscape # generates what is missing decktalk soundscape --only music # one item; --force regenerates it ``` ```text theme={null} == music -> /private/tmp/dtclip0/build/music/underscore.mp3 [planned] POST https://api.elevenlabs.io/v1/music {'prompt': 'calm, minimal instrumental underscore for a lesson: soft piano and warm pads, slow, no drums, no melody hooks, unobtrusive under a narrator', 'force_instrumental': True, 'model_id': 'music_v2', 'music_length_ms': 120000} ``` Music longer than `max_music_chunk_seconds` is requested in chunks and joined with a crossfade. Every output has a manifest with its request hash, so a second run sends nothing unless a prompt changed. These calls spend ElevenLabs credits, and the dry run shows exactly what would be sent. [decktalk.toml](/reference/decktalk-toml#soundscape) lists every key of the three tables. ## Bringing your own Point `mix.underscore`, `mix.ambience`, and each `[[mix.sfx]]` at any audio file that ffmpeg can read. Keep a file of your own at a path that no `[soundscape]` table writes to, because `decktalk soundscape` regenerates any output that has no manifest beside it and would overwrite the file. The scaffold's `[soundscape.music]` writes to the `mix.underscore` path, so either remove that table or point `mix.underscore` at your own file and give the table its own `out`. [Put a sound effect on a cue](/guides/sfx-on-a-cue) walks through one effect from prompt to mix.