Pro-Voice (Add-on)

Studio‑grade realism trained only on your recordings—no shared data, no guesswork.

What it is

Pro Voice builds a dedicated voice model from your recordings - exclusively your voice. Pro Voice learns solely from your speech for a richer, more authentic sound. They have much much more fine tuning and use significantly greater sample lengths to achieve a hyper realistic outcome.

Highlights

  • Exclusively yours: Model trained only on your audio.

  • Richer detail: Provide ≥30 minutes (ideal ~60 minutes) to capture nuances.

  • Studio polish: Advanced noise reduction & high‑fidelity synthesis for broadcast‑quality calls.

  • Consistent over time: Long calls and streams keep the same clear tone start‑to‑finish.

Not quite right even after setup? Email [email protected] and we’ll partner with you to dial it in.


Why upgrade

  • Exact match to your natural voice & accent

  • Higher clarity in noisy listening environments

  • Consistency across long sessions and varying prompts


Requirements

  • Audio duration: Minimum 30 minutes; 60 minutes recommended.

  • Quality: One clean session in a quiet space (same mic, same room, steady tone).

  • Format: WAV/MP3 preferred.


How to upgrade

  1. Go to Studio → Voice and click settings gear icon (top‑right).

  2. In the Pro Voice banner, click Upgrade Now.

  3. Click Continue to Payment and complete checkout ($150 / month).

  4. After payment, email [email protected] with your training audio:

    • Record or upload at least 10 minutes (30 minutes ideal).

    • Follow the recording guidelines below.

    • Include your Delphi handle and any pronunciation notes.

We’ll handle training and notify you when your dedicated model is live.


Recording guidelines (capture once, capture well)

  • Room: Silent, low‑echo space; turn off HVAC, devices, notifications.

  • Mic: Prefer XLR mic + interface (e.g., AT‑2020 or Rode NT1 with Focusrite). USB is OK; avoid Bluetooth and Zoom/call captures.

  • Distance: ~2 fists from the mic with a pop filter.

  • Levels: Aim −23 dB to −18 dB RMS, peaks < −3 dB.

  • Performance: Steady pace, single speaker, one language.

  • Editing: Trim long silences & filler words if you want a polished tone.


What to read (suggested 30‑min script plan)

  • Intro & bio (3–5 min): Who you are, domains, typical questions.

  • Explainers (10–12 min): Teach 3–4 concepts in your natural style.

  • Q&A (10–12 min): Answer common questions out loud.

  • Pronunciation list (2–3 min): Proper nouns, names, brands; include variants.


After training

  • Test in Voice Playground, then place a live call for a true check.

  • Fine‑tune Stability / Similarity / Speed.

  • Use Custom Pronunciations for tricky names.

  • Re‑record and retrain only if your gear/space changes significantly.


FAQs

  • How long does training take? We’ll email when it’s ready (timing depends on queue and audio length/quality).

  • Can I send multiple clips? Yes—but record them with the same mic/room for consistency. A single session is best.

  • What about accents? Pro Voice captures your natural accent more faithfully than defaults.

  • Can I downgrade later? Yes—disable the add‑on on the Add‑ons page; your default voice settings remain.

  • Is my audio shared? No—Pro Voice models are trained only on your submitted recordings.


Pre‑submit checklist

Last updated