Every feature,
in the order you'll meet it.

Paste or type a script. The import screen ships with a collapsible format guide and a built-in sample script for first-time users.

The parser detects speakers, SFX, and music cues across colon (`HOST:`), bracket (`[Host]`), parenthesis (`(Host)`), and standalone-name formats, and assigns confidence scores to each attribution.

Every detected turn shows its speaker with a green / orange / red confidence indicator. Lines that need attention are highlighted; the rest you scroll past.

Add new speakers, merge duplicates, batch-assign unreviewed turns, and give each character a unique color badge for visual distinction.

Edit the raw script after import. An audio-tag reference panel makes inserting SFX and music cues quick. Re-parsing replaces existing speaker assignments and generated audio (with a confirmation warning).

Browse the ElevenLabs catalog with search, category and type filters, and paginated results. Inline play / stop preview on every row.

Map each speaker in your script to a TTS voice. The mapping screen shows assignment status for every speaker with a clear progress indicator.

Independent sliders for stability, similarity boost, style exaggeration, and speed, plus a speaker boost toggle. V3 models show a simplified panel.

Per-voice metadata, verified languages, category, and a `Find Similar` feature that analyzes the voice's preview audio and returns related options from the catalog.

Per-project rules for tricky names and technical terms. Alias replacements or phoneme overrides in IPA or CMU Arpabet. Synced to ElevenLabs before generation.

Each speech turn is generated with the assigned voice profile and contextual parameters (previous / next text, request-ID history) for voice continuity. SFX and music turns hit dedicated endpoints.

Batch multiple speech turns into a single multi-speaker call for natural conversational flow. The app respects a 2,000-character and 10-unique-voice limit per batch and splits larger scripts automatically.

Write `[SFX: door slam (2s)]` directly in your script. flexCast generates the effect with a duration clamped between 0.5 and 30 seconds.

Write `[Music: upbeat jazz intro (10s)]` and the app generates it. Duration clamps between 3 seconds and 10 minutes; an instrumental-only flag is available.

An animated circular progress ring shows percentage, current turn, estimated time remaining, and elapsed time. Cancel at any point — already-generated audio is preserved.

If generation is interrupted or cancelled, the app detects which turns already have audio and offers a one-tap resume that skips completed turns.

Every generated file is checked for corruption, silence, and minimum duration. Files with detected issues are flagged with a warning badge in post-production.

With no API key configured, a mock TTS service returns silent WAV audio with realistic durations. Every screen — import, review, voices, generate, edit, mix, export — works end to end.

Play each segment individually. Simple and advanced modes are available — advanced shows per-turn controls, variant counts, and quality indicators.

Swipe a turn or use its context menu to regenerate it. The new take is saved as an additional variant — your previous take is never overwritten.

Each turn can have multiple takes. Browse them, play them back to back, mark one as active, and delete the rest. `Keep Only Active` cleans up across the whole project.

Two levels of control: a project-wide default, and per-turn custom pauses dialed in with a slider from 0 to 2 seconds.

Exclude individual turns from the final mix without deleting them. Excluded turns appear with strikethrough text and are skipped during mixing — toggle them back any time.

Edit a turn's dialogue text directly from post-production. After saving, optionally regenerate the audio immediately to match.

Concatenate every active, non-excluded turn into a single M4A / AAC file with configurable pauses between segments.

Listen to the mixed output with play / pause, skip forward and back 15 seconds, and a waveform scrubber for precise seeking.

Stream the first ~2,000 characters of dialogue with assigned voices via the ElevenLabs dialogue API. A segment timeline shows colored bars per speaker.

Share the final M4A via the iOS share sheet — Files, AirDrop, email, or any app that accepts audio.

Create, rename, duplicate, and delete projects. The list is searchable and sorted by most recently updated.

Script. Voices. Generate. Production. A progress bar tracks where you are; tabs unlock as the project advances; the current step is auto-selected based on status.

Deep-copy a project including script, turns, speakers, voice profiles, and pronunciation rules. Audio is intentionally not copied — the duplicate is a fresh starting point.

A pre-populated sample script is one tap away from the empty-state screen.

Enter and store an ElevenLabs API key in the iOS Keychain. Remove it any time. The app clearly displays whether it's running in API or demo mode.

Test the saved key by fetching available voices. Pick the default TTS model from Multilingual v2, Turbo v2.5, Turbo v2, English v1, or v3.

Light taps for selections, medium impacts for state changes, ticks for slider adjustments, notification haptics for generation milestones and errors.

Ephemeral feedback at the top of the screen for assignments, duplications, regeneration completion, and errors. Four styles, matching icons, VoiceOver-accessible.

If the local SwiftData store is corrupted on launch, the app attempts recovery in three tiers: normal open, delete-and-retry, and in-memory fallback. Users are notified of any data reset.