Auto-Cast in one tap, or run the room yourself.
The voice mapping screen lists every speaker in your script alongside their cast status. Hit Auto-Cast All and flexVox analyzes each speaker's role, queries the ElevenLabs catalog, and assigns the best-match voice. Manual assignments are preserved.
Want more control? Open the library as a casting session. It stays open until you tap Done. A progress counter ("3 of 5 cast") tracks every decision, and the session auto-advances to the next unvoiced speaker as you commit each choice. You can jump back to any speaker at any time to audition alternatives or swap voices.
The session is the unsung hero. Without it, casting feels like paperwork. With it, casting feels like directing.
The AI archetype banner shows its work.
Inside the casting session, an inline banner shows the AI-recommended archetype for the current speaker and explains its reasoning — "close-mic confidant · warm · slightly raspy." You can take the suggestion, search for something close, or ignore it entirely. The banner never picks a voice for you; it just tells you what it would pick, and why.
- What it readsEvery line that speaker says, plus character profile (if set).
- What it returnsAn archetype phrase + a short reasoning sentence.
- What it never doesOverride a voice you've already assigned.
Per-speaker tuning, collapsed by default.
Once a voice is assigned, you can dial it in. flexVox exposes the four ElevenLabs parameters for every speaker — independently — with a speaker boost toggle on supported models. v3 models get a simplified panel with only what matters for that generation.
- StabilityHow consistent the voice is across runs. Lower = more expressive.
- Similarity boostHow closely the output tracks the original sample.
- StyleExaggeration of the voice's natural character.
- SpeedPlayback speed multiplier. Defaults to 1.0.
- Speaker boostExtra clarity for difficult listening environments (supported models only).
The fine-tune section is collapsed by default. We don't want a wall of sliders staring at you when all you wanted was to assign a voice.
Cast versioning. Re-cast freely.
flexVox auto-saves a snapshot of your voice assignments before every generation run. You can also save manual snapshots with custom names — a "before the rewrite" snapshot, a "what if we made ALEX female" snapshot. Auto-snapshots are pruned to the most recent ten.
Restoring a snapshot applies its voice mappings back to the matching speakers in the project. Cast versioning is a Studio feature.
And when you do re-cast someone, every previously-generated turn of theirs shows a RECAST badge in post-production. Tap Regenerate Changed to update them all at once — or leave it alone and only refresh the lines that need it.
What's inside the library.
- Search & filter By name, category (premade, cloned), and voice type. Paginated.
- Inline preview Play / stop button on every voice row. Auditions are instant.
- Voice detail Description, metadata labels, verified languages, category.
- Find similar Analyzes the current voice's preview audio and returns related voices.
- Pronunciation dictionary Per-project text aliases or phoneme overrides (IPA / Arpabet). Studio.
- Voice change detection Every audio asset tracks the voice that produced it. Recasts get badged.