Settings - Reasonate User Guide

Tab 1General

Settings General tab: API key with Test and Save buttons, cache statistics and size limit, appearance and diagnostics toggles, About section with support buttons — General: key, cache, appearance, About.

Setting	What it does
ElevenLabs API key	Test connection verifies the key; Save & fetch voices stores it and loads your voice library. The key lives in REAPER's local settings - never in project files.
Cache	Shows the current project's cache (file count, size, location). Cache size limit (GB) caps it - oldest unused files are evicted first; 0 means unlimited. Clear cache wipes it (next renders will be paid fresh calls).
Decorative background	The subtle gradient behind the interface. Turn off for a flat look.
Diagnostic logging	Writes technical details to the REAPER console. Leave off unless you are chasing a problem (see Help).
About	Version, license, the support links (also under the ♥ button in the header) and Check for updates - compares your version against the latest release and links to the releases page. Reasonate also checks quietly once a day and shows a small note in the header when an update exists; it never interrupts you with popups, and installing updates stays in ReaPack's hands.

Tab 2AI

Settings AI tab: provider override, six LLM provider sections with API key and model, and per-feature model overrides — AI: your LLM keys and which feature uses which model.

Three features use a large language model: Dubbing translation, TTS Enhance and SFX scene analysis. They all run on your own keys.

Setting	What it does
Provider override	Which configured provider is the default. Auto picks the first one with a key.
Provider sections	Anthropic Claude, OpenAI, Google Gemini, DeepSeek, xAI Grok and Mistral (EU) - paste a key, press Test to verify it (a free check, no tokens spent), pick a model. Configure one or several.
Per-feature model	Give each feature its own provider and model (for example: a top model for dubbing translation, a cheap one for SFX ideas). Default follows the global provider.

Keys are stored like the ElevenLabs key: locally, never in project files, and passed to the network layer via protected files rather than command lines.

Tab 3TTS

Settings TTS tab: advance cursor toggle, dialogue split per speaker, output format and all-time character counter — TTS: cursor behavior, output format, billing counter.

Setting	What it does
Advance edit cursor	After Generate, jump the REAPER cursor to the end of the new item - write, generate, keep writing.
Split into per-speaker tracks	After a dialogue Generate, detect the speakers in the result and lay each on its own track (one extra speech-to-text call per Generate).
Output format	MP3 44.1 kHz at 128 or 192 kbps, or uncompressed PCM (higher tiers). Applies to TTS renders.
Billing counter	All-time TTS characters generated, with a reset button. Cache hits are not counted - they are free.

Tab 4Voice Replacement

Settings Voice Replacement tab: output layout selector and recording options for input channel, pre-roll and monitoring — Voice Replacement: where AI tracks go, how recording behaves.

Setting	What it does
Output layout	Folder child nests the [AI] track inside the source track's folder (default); flat sibling puts it next to the source. A migrate button converts old projects to the folder layout.
Recording	Input channel (mono 1 / mono 2 / stereo), pre-roll countdown (0-5 s) and input monitoring for the in-panel ● Record feature.

Tab 5Repair

Settings Repair tab: auto-match speaker pace toggle and ripple all tracks toggle, each with an explanation — Repair: pacing and timeline shifting.

Setting	What it does
Auto-match speaker pace	Off by default: the voice and sentence decide the pacing naturally, and you can nudge speed manually per edit. On: Reasonate measures the take's pace and generates to match, keeping the better of two renders.
Ripple all tracks	When an edit changes the item's length, shift downstream material on all tracks (on) or only the edited track (off, default - safe when other tracks are not synced to the dialogue).

Loudness matching is always on - patches are level-matched to the surrounding words automatically.

Tab 6Dubbing

Settings Dubbing tab: default TTS model and style preset, dub track placement, speaker separation, alignment and stretching toggles, prompt caching and cost alert threshold — Dubbing: project defaults and timing behavior.

Setting	What it does
Default TTS model / style preset	Prefills the Start Project modal.
Dub track placement	Dub tracks as folder children of the source, or as flat tracks below it (source fader then does not affect the dubs). Applies to newly created tracks.
Speaker separation	How eagerly speaker detection decides "this is a different person". Leave on auto unless detection merges or splits people.
Auto-run Forced Alignment	After each Generate Dub, compute per-word timing of the result (needed for per-word lip-sync and precise playback mapping).
Voice Isolator pre-clean	Default state of the pre-clean option for new projects.
Stretch dub items to segment span	Fit each dub item into its original segment's slot to prevent overlap, with a bypass for very short segments.
Per-word lip-sync	Experimental: elastic stretch markers align each word to the original word timing.
Sliding context window	The translator sees neighboring segments (2 previous + 1 next) for coherent phrasing across cuts.
Anthropic prompt caching	90% off repeated context when translating with Claude. Keep it on.
Cost alert threshold	Generate Dub asks for confirmation when the projected spend crosses this amount.