← Back to voxxenstudio.com

Voxxen User Manual

Everything you need to go from raw manuscript to multi-character audio.

What is Voxxen?

Voxxen is a desktop recording studio built for authors, narrators, and playwrights. You record every character in your own voice — then Voxxen converts each line into a distinct character voice using its built-in RVC voice engine. The result is a set of clean, per-character WAV stems ready to mix in your DAW.

There is no AI text-to-speech. Your performance is always the source. The timing, emotion, and pacing you put into a take carries through. Voxxen only swaps the timbre — who is speaking — not how they speak.

Voxxen is a character creation tool, not a voice cloner. The 70 included voice models are neutral raw material — each one provides a gender, age, and accent. What makes a character is the combination of your performance and your tuning choices. Two narrators using the same voice model will sound completely different from each other because the creative decisions are different. The resulting voice belongs to your character, not to the original speaker.

When you find a combination that works, save it to your Character Library. A character you built for one audiobook can be added to the next project in a single click, with all settings intact.

Voice Profile
→
Record
→
Convert
→
Export Stems
→
Mix in DAW

Voxxen works fully offline after installation. No subscription, no cloud processing, no internet required to record or convert.

System Requirements

  • OS: Windows 11 (64-bit)
  • GPU: NVIDIA GPU with CUDA support — GTX 1060 or newer recommended. RTX series strongly preferred for conversion speed.
  • VRAM: 6 GB minimum, 8 GB+ recommended
  • RAM: 16 GB recommended
  • Disk: ~20 GB download, ~27 GB installed
  • Audio: Any USB or XLR microphone. A USB audio interface is recommended for best quality.
No NVIDIA GPU? In v1, GPU-accelerated conversion requires an NVIDIA GPU (CUDA). AMD and Intel GPUs fall back to CPU mode — conversion still works but is significantly slower (minutes per segment rather than seconds). Broader GPU support is planned for a future update as PyTorch's AMD and Intel backends mature on Windows.

Installation

  1. Download VoxxenSetup.exe from the Download button on the Voxxen website.
  2. Run VoxxenSetup.exe. If Windows shows a SmartScreen warning, click More info → Run anyway. This is normal for new software.
  3. The installer downloads the full Voxxen package (~20 GB). Leave it running — this will take a few minutes depending on your connection.
  4. Once downloaded, the Inno Setup wizard launches. Follow the prompts and choose an install location with at least 27 GB of free space.
  5. After installation completes, launch Voxxen from your desktop shortcut or Start Menu.
Tip — Install on your fastest drive. Voice models load from disk during conversion. An SSD (especially NVMe) gives significantly faster conversion start times than an HDD.

Activation & Trial

When you first launch Voxxen, you will be prompted to activate.

Free Trial

Click Start Free Trial to begin your 30-day trial. You will be asked for your email address — this is optional, but providing it lets us send you a reminder before your trial expires and a discount offer if you decide to purchase.

The trial is fully functional. Every feature available in the paid version is available during your trial.

License Key

If you have purchased a license, click Enter License Key and paste your key. Keys are delivered by email immediately after purchase. Your key is tied to one machine. If you need to transfer your license to a new computer, contact support.

Your trial never resets. The trial clock is tied to your machine, not to the app files. Reinstalling or deleting settings does not reset it.

STEP 1 Your Voice Profile

The first time you launch Voxxen, you will be asked two questions about your performing voice. This takes about ten seconds and only happens once.

Why it matters

Voxxen's voice models are trained on a wide range of voices — male, female, higher, lower. When you record a chapter, your voice is the source signal that gets converted into each character. The further apart your voice and the target model are in pitch and register, the more the tuning panel needs to compensate.

Your voice profile tells Voxxen how to set the starting defaults for every character's tuning sliders — Pitch, Formant Ratio, Index Rate — so the first conversion attempt is already in the right ballpark without any manual adjustment.

The two questions

  • I perform as a… Male or Female. This sets the baseline pitch reference for conversion.
  • My pitch is… Higher than average / Average / Lower or deeper. This fine-tunes the defaults for voices at the extremes of each gender range.
You can change this any time. Go to Settings → Performer Profile to update your profile. Changes apply new defaults to characters that haven't been individually tuned yet. Characters you have already saved custom tuning settings for are not affected.

Individual character tuning can always be overridden manually in the Voice Tuning panel — the profile just gives you a sensible starting point so you are not tweaking every character from scratch.

STEP 2 Create a Project

From the sidebar, click + New Project. You will be asked to choose a project type:

Audiobook

For novels, non-fiction, or any long-form written work. You create chapters manually and record each one. Import a PDF of each chapter's text to auto-populate the text bubbles.

Audio Play / Script

For screenplays and stage plays formatted in standard US screenplay format. Import your full PDF once and Voxxen automatically splits it into three acts based on page count. Each act becomes a recordable chapter in the Studio.

Tip — Screenplay PDFs. Voxxen's screenplay parser requires standard US screenplay format — ALL-CAPS character cues, standard margins. PDFs from Final Draft, Fade In, or Highland export cleanly. Scanned PDFs are not supported. See PDF Tips.
Tip — Audiobook PDFs. For audiobooks, any text-based PDF works — Adobe Acrobat, Word-exported, InDesign, or any app that preserves selectable text. The only requirement is that you can highlight and copy text from the file. Scanned documents (images) will not parse.

STEP 3 Add Characters

Before recording, build your cast. Every speaking role — including the Narrator — needs a character entry with a voice assigned.

  1. Go to the Characters page from the sidebar.
  2. Click + Add Character. Enter the character's name, age, and gender — these are notes for your reference only and do not affect conversion.
  3. In the Also known as field, enter every other name, nickname, or title the author uses for this character in the text — separated by commas. For example, a character named Dr. Michael Brown might be called Doc, Mike, or Dr. Brown in dialogue. Voxxen's PDF parser checks all of these when attributing lines to the right character.
  4. Assign a Voice from the dropdown. Voxxen includes 70 unique trained voice models. Listen to previews to find the right match for each role.
  5. Pick a color for the character. Colors appear on text bubbles in the Studio so you can see at a glance who is speaking.
  6. Repeat for every character in your project, including the Narrator.
Tip — Also Known As is the single most impactful field for prose parsing. Novels rarely call characters by one consistent name. A character may be Dr. Michael Brown in formal scenes, Dr. Brown to colleagues, and Doc or Mike to close friends. The PDF parser can only attribute dialogue correctly if it knows all of those names belong to the same character. Screenplay dialogue is different — character cues are always the single canonical name in ALL CAPS, so the Also known as field matters much less there.
Tip — Cast strategically. Match vocal age and gender of the voice model to your character concept. The conversion preserves your performance best when the source voice (yours) and the target voice are in a similar register. A large pitch difference can be tuned in the Voice Tuning panel.

You can add or edit characters at any time, even after recording has begun.

STEP 4 Chapters & Acts

Audiobook — Adding Chapters

  1. Go to the Chapters page.
  2. Click + Add Chapter for each chapter in your book. Give each one a name and number.
  3. Click Open Studio on a chapter row to enter the Recording Studio for that chapter.

Audio Play / Script — Importing Acts

  1. Go to the Acts page.
  2. Click Upload Script and select your screenplay PDF.
  3. Voxxen parses the PDF and presents a split dialog. Review where Acts 1, 2, and 3 begin and adjust the page boundaries if needed.
  4. Confirm — three Act entries are created automatically.
  5. Click Open Studio on any act to begin.
Wrong PDF? Use the Clear Script button inside the Studio to remove a mis-imported PDF without deleting your chapter or any recordings.

STEP 5 The Recording Studio

The Studio is where recording happens. It is split into two panels: the Script panel (left) and the Controls panel (right).

Import Your PDF

Click Import PDF in the toolbar to load your chapter or script text. Voxxen parses the PDF into text bubbles — one per line of dialogue or narration — and assigns each bubble to the character whose name appears on the cue line above it.

Review the bubbles and correct any mis-attributed lines by clicking the character name on a bubble and selecting the right one. You can add, delete, or reorder bubbles manually using the toolbar buttons.

Tip — Prose authors: fill in "Also Known As" before importing. If a character appears in your cast as Dr. Michael Brown but the text refers to him as Doc, Mike, or Dr. Brown, the parser won't connect those lines to him and will fall back to Narrator. Go to the Characters page, edit each character, and add every name or nickname the author uses in the Also known as field before running Import PDF. This one step prevents most attribution errors in prose chapters.
You don't need a PDF to record. If you prefer to record freely without a script, skip the import and just click 🎙 Record. You can add text bubbles manually or leave the script panel empty.

Set Up Your Microphone

Before recording, configure your input in the Controls panel:

  1. Under INPUT DEVICE, select your microphone from the dropdown. If you are using a USB mic or audio interface, it should appear here by name.
  2. Enable Studio Input Mode. This activates WASAPI Exclusive mode, bypassing all Windows audio processing — noise reduction, loudness normalization, EQ, and echo cancellation. Without it, Windows boosts and compresses the signal before Voxxen sees it. RVC is trained on raw, dry audio; a processed waveform causes the converted voice to sound mechanical, with truncated consonants and unnatural cadence.
  3. Watch the INPUT LEVEL meter while speaking at your normal recording volume. Aim for green — if it hits red, your audio is distorting. Back off from the mic slightly.
If the INPUT LEVEL meter reads low even at normal speaking volume: Open Windows Sound settings → Input → select your device → raise the recording level slider. The right value varies by microphone and interface — there is no universal percentage. Use the Voxxen meter as your guide. Some devices default very low on first connection, and a small adjustment can make a large difference in signal strength.
Always use Studio Input Mode. Windows audio processing makes your recording sound fine to your ear — that's the problem. RVC sees a compressed, normalized signal instead of clean dialogue, and the conversion goes wrong: mechanical tone, clipped consonants, unnatural rhythm. Note: some laptop sound cards enforce noise reduction in hardware regardless of Windows settings. An external USB mic or audio interface sidesteps this entirely and is the recommended setup.

Recording a Chapter Take

Voxxen records your entire chapter (or act) as one continuous take. You perform all characters in your own voice, reading the text bubbles in sequence.

  1. Click 🎙 Record to start. The timer begins and the level meter goes live.
  2. Read each bubble aloud. Click → Next or press Space to advance to the next bubble as you go — this marks your timing for later conversion.
  3. Made a mistake? Press ← to go back one bubble, then continue reading from that point. The timing marker moves with you.
  4. When finished, click ■ Stop. Your take is saved automatically.
  5. Click ▶ Play to review the full take.
Performance tip. Read every character in your own natural voice — just vary your pacing and energy to differentiate them. Voxxen handles the timbre change. Do not strain your voice trying to impersonate the character; the conversion will do that for you.

STEP 6 Voice Tuning

The Voice Tuning panel lets you dial in the conversion settings for each character before committing to a full chapter conversion. Think of it as a quick audition tool.

The Sliders

  • Pitch — Shifts pitch up or down in semitones. Use this if a converted voice sounds too high or too low for the character. Most voices need little or no pitch adjustment.
  • Formant Ratio — Adjusts the vocal tract size. Higher values make the voice sound younger or smaller. Lower values add depth. Default is 1.14.
  • Index Rate — Controls how strongly the voice model influences the output. 0 = sounds like you; 100 = full character voice. Start at 75 and adjust to taste.
  • Gain (dB) — Output volume trim. Use this to match loudness across characters so you are not constantly adjusting your DAW faders.

Testing Your Settings

  1. Click ● Rec to record a short 5-second test clip from your microphone.
  2. Click ▶ Convert to convert the clip with the current slider values.
  3. Listen to the result. Adjust sliders and convert again until you are happy.
  4. Click Save to store these settings for the active character. They will be used automatically when you convert the full chapter take.
Tune each character separately. Select a different character from the character list, then tune and save. Each character stores its own settings.

STEP 7 Character Library

The Character Library is your personal roster of finished character voices. Once you have tuned a character and are happy with how they sound, you can save them — with their name, role, archetype, and all tuning settings — and bring them into any future project in one click.

What a character is

A character in Voxxen is not a voice model. The voice model is just the starting point — like choosing a string instrument before you play it. A character is the combination of:

  • The voice model you selected (providing gender, age, and accent)
  • Your tuning settings (Pitch, Formant Ratio, Index Rate, Gain)
  • The creative choices you made — which model felt right for this role, and how you pushed the sliders

Two people using the same voice model with different tuning produce completely different characters. The character belongs to your creative work, not to the voice model.

Saving a character to your library

  1. Open the Recording Studio and select the character you want to save from the character list on the right.
  2. Tune the character using the Voice Tuning panel until you are happy with the result. Record a short test clip and convert it to confirm.
  3. Click 🎭 Save to Library in the tuning panel.
  4. Fill in the character's details: name, occupation, archetype (e.g. "The Mentor", "The Villain"), age, gender, and any notes. These are for your reference — they appear on the character card in the library so you can find and recognise characters at a glance.
  5. Click Save. The character and all their settings are stored in your personal library.
Your raw performance and converted preview are both saved. When you save a character, Voxxen stores the short test clip you recorded as well as the converted result. From the library, you can play back both — the raw performance (so you know what kind of vocal delivery works for this character) and the converted voice (to audition them quickly before adding to a project).

Adding a library character to a new project

  1. Open the Character Library page from the sidebar.
  2. Find the character you want. Use the ••• menu on their card.
  3. Click Add to Cast. The character is added to the current project's cast with all their tuning settings pre-loaded.
  4. You can rename them for this project (e.g. "Winona" in the library becomes "Julia" in this book's cast) without affecting the library entry.

Updating a library character

If you retune a character while working on a project, those changes are saved to that project's cast only — they do not automatically update the library entry. If you want to update the library version, click 🎭 Save to Library again. Voxxen will ask whether to update the existing library entry or save as a new one.

Your library is not voice cloning. Library characters are original creative works. They combine a neutral voice model with your performance choices and your tuning decisions. The result does not reproduce any specific person's voice — it is a fictional character that exists because you built it.

STEP 8 Voice Conversion

Once you have a recorded chapter take and your characters are tuned, convert the audio.

Converting a Chapter

  1. With a take recorded, click ▶ Convert in the main transport.
  2. Choose to convert all characters or select specific ones.
  3. Voxxen splits the take at your marked bubble boundaries and runs each segment through the assigned voice model.
  4. Converted takes appear in the takes list with a ✓ indicator. Click any to preview.

Conversion Time

On an RTX 3070 or newer, expect roughly 1–3 seconds per short dialogue segment. A full 30-minute chapter take may take 5–15 minutes to convert fully, depending on your GPU and the number of characters.

Re-converting. You can re-record a chapter take and re-convert at any time. Previous conversions are replaced. Nothing is permanent until you export.

STEP 9 Export Stems

Click Export Stems to export your converted audio as individual WAV files — one per character.

Files are saved to your project's output folder and named by character:

  • Chapter_01_Narrator.wav
  • Chapter_01_Alex.wav
  • Chapter_01_Maria.wav
  • …and so on

Import these stems into Logic Pro, Audition, Cubase, Ableton, or any DAW for final mixing, noise reduction, room treatment, and mastering. Voxxen is the first step — your DAW is the last.

Sample rate. All Voxxen voice models are trained at 48 kHz. Export stems are 48 kHz / 24-bit WAV. Most DAWs default to 44.1 kHz — set your DAW project to 48 kHz to avoid unnecessary resampling.

Recording Best Practices

Room Setup

  • Record in the quietest space available. HVAC noise, fans, and street sound will be captured by the mic and can degrade conversion quality.
  • Soft furnishings (carpet, curtains, bookshelves) naturally reduce room echo. A wardrobe full of clothes is a surprisingly good recording booth.
  • Avoid recording directly in front of a hard wall. Angle your mic slightly off-axis from reflective surfaces.

Microphone Technique

  • Get close. Position the mic 2–4 inches from your mouth. The closer you are, the stronger and cleaner the signal — RVC converts better with a strong direct signal than a distant one padded with room noise. Find a position and stay there; distance variation between lines causes noticeable volume inconsistency.
  • Angle slightly off-axis. Point the mic just past the corner of your mouth rather than straight at your lips. This reduces plosives — the burst of air on hard P and B sounds that can distort the recording.
  • Use a pop filter. A pop filter (foam windscreen or mesh screen on a gooseneck) is cheap and eliminates plosive spikes that are impossible to fix after recording. Highly recommended.
  • Treat the room. Record in the quietest space available. Soft furnishings (curtains, carpet, a closet full of clothes) absorb reflections. Hard, bare walls create a roomy sound that carries through conversion.

Recommended Microphone Setup

  • External USB mic or XLR interface. A dedicated external microphone gives you a clean, consistent signal and bypasses any hardware noise reduction baked into laptop sound cards. This is the single biggest hardware upgrade you can make.
  • Dynamic mic (recommended for home use). Dynamic mics (like the Shure SM58, PG58, or Audio-Technica AT2005) reject background noise naturally and are forgiving of untreated rooms. They work best close to the mouth.
  • Condenser mic (studio environments). Condensers are more sensitive and capture detail well, but they pick up everything — room noise, HVAC, keyboard clicks. Best used in a treated space.
  • Avoid laptop or webcam mics. Built-in mics are usually subject to hardware-level noise processing that cannot be bypassed, even with Studio Input Mode enabled. The signal they produce is rarely clean enough for consistent conversion quality.

Your Performance

  • Read naturally. The RVC engine preserves timing, rhythm, and emotional inflection. A flat, "announcer-voice" read will produce flat conversions.
  • Pitch — you don't need to match the character's pitch. The Voice Tuning panel handles pitch shifting, so record at your natural speaking pitch and adjust from there.
  • Accent — you do need to perform it. RVC converts timbre, not phonetics. If you assign an Irish or Scottish voice, the accent in the preview comes from the original speaker — it will only carry through in your conversion if you attempt that accent while recording. A flat performance into an accented model produces a flat result. It may feel unnatural to perform an accent into raw audio, but the conversion smooths it into the target voice convincingly.
  • Read continuously without stopping for small mistakes. Use the back-bubble button ← to re-mark the line and continue. You can clean up timing in the Studio later.

PDF & Script Tips

What Makes a Good PDF for Voxxen

  • Text-based, not scanned. Voxxen reads the text layer of a PDF. Scanned documents are images and cannot be parsed. If you can select and copy text from the PDF, it will work.
  • Standard screenplay format (for scripts). Character cues must be in ALL CAPS and follow standard US screenplay margins. PDFs from Final Draft, Fade In, or Highland export cleanly.
  • Clean chapter files (for audiobooks). A separate PDF per chapter works best. Long documents covering many chapters may slow parsing and increase misattributions.

Fixing Mis-attributed Bubbles

If the parser assigns dialogue to the wrong character, click the character name on any bubble to reassign it. You can also split or merge bubbles using the toolbar buttons. These edits are saved with the project.

If the Import Fails

Use Clear Script in the toolbar to remove the import and try a different PDF. Your recordings are not affected.

Voice Quality Tips

Getting the Best Conversion

  • A clean, dry mic signal converts better than a processed one. Studio Input Mode is the single most important setting.
  • Stay in the green zone on the level meter. Distorted (clipped) audio cannot be fixed in conversion and will produce artifacts.
  • If a converted voice sounds "watery" or has strange artifacts, try reducing Index Rate from 75 to 60. This blends more of your original voice into the output and often cleans up problem areas.
  • Pitch shifts above +6 or below −6 semitones can cause quality loss. If you need a large pitch change, consider choosing a different voice model that is naturally closer to your target.

Matching Voices to Characters

  • Choose voice models whose natural age and gender roughly match the character. The engine performs best when source and target are in a similar register.
  • Use the Voice Library page to preview all 70 models before assigning.
  • A character can be re-assigned a different voice at any time — including after recording. Re-convert the chapter to hear the new voice.

Frequently Asked Questions

My trial expired. Can I still access my projects?
Yes. Your projects, recordings, and settings are all stored on your machine and are not deleted when the trial ends. Purchase a license and activate to resume recording and conversion.
Can I use Voxxen on more than one computer?
A license is tied to one machine. Contact support if you need to transfer your license to a new computer.
Can I add my own voice models?
Not in the current version. Voxxen ships with 70 curated, professionally trained models. Additional voice packs are planned for a future update.
Does Voxxen work without an internet connection?
Yes, fully, after initial installation and activation. Recording, conversion, and export all run entirely on your local machine.
What file formats can I export to?
Voxxen exports 48 kHz / 24-bit WAV stems. Convert to MP3 or AAC in your DAW or with a free tool like Audacity after mixing.
I recorded a great take but the conversion sounds wrong. What do I do?
First, check that Studio Input Mode was enabled during recording. Windows audio processing (noise reduction, loudness normalization) corrupts the signal RVC needs — the result is mechanical tone, clipped consonants, or unnatural cadence. Artifacts from processed audio cannot be corrected; re-record with Studio Input Mode on. If you're on a laptop and the problem persists, your sound card may enforce noise reduction in hardware — switch to an external USB mic. If the recording was clean, try re-tuning: reduce Index Rate slightly and adjust Pitch.
Can I record multiple takes and choose the best one?
Yes. Record as many takes as you like — each one is saved. Review takes in the takes list and mark the one you want to convert as the active take.
What is the difference between a voice model and a character?
A voice model is neutral raw material — it provides a gender, age, and accent profile. A character is what you build on top of it: your tuning choices (pitch, formant, index rate), your performance, and the creative decisions you made about how this role should sound. Two people using the same voice model will produce entirely different characters because those creative decisions differ. The voice model is the instrument; the character is your performance on it.
Is Voxxen voice cloning? Is it legal?
No. Voice cloning typically means replicating a specific identifiable person's voice without their consent. Voxxen does not do this. The 70 included voice models are trained on licensed, royalty-free datasets (Mozilla Common Voice, VCTK, and LibriTTS — all under Creative Commons licenses) and are designed to produce distinct fictional character voices, not to imitate any individual. The output of Voxxen is a character voice created by you — your performance combined with your tuning choices and your selection of a neutral starting model. This is the same creative process a voice actor uses to develop a character voice for a role. Voxxen is a production tool for authors and narrators, not a voice impersonation tool.
Can I save a character and use them in multiple projects?
Yes — that is exactly what the Character Library is for. Tune a character in the Voice Tuning panel, click 🎭 Save to Library, and give them a name, role, and archetype. They are stored in your personal library with all their settings. Open any project, go to the Character Library page, and click Add to Cast to bring them in — all tuning pre-loaded, ready to record.
Can I rename a character in one project without changing them in the library?
Yes. A character's name in your cast is separate from their library entry name. If the library entry is "Winona — The Girl Next Door" and you add her to a project as "Julia", renaming her in that project only affects that cast — the library entry stays as Winona. The link between the cast member and the library entry is shown under their name on the cast card, so you always know which library character you are working with.

Troubleshooting

"Voxxen could not locate the voice engine."
The voice engine folder has been moved or the installation is incomplete. Reinstall Voxxen to restore the engine to the correct location.
The PDF import button is grayed out.
The PDF import button is only available when you have at least one character in your cast. Go to the Cast page, add your characters, then return to the Studio and try again. If you have characters and the button is still unavailable, try reinstalling Voxxen.
Conversion is extremely slow.
Make sure your NVIDIA GPU is active and not in a power-saving state. On laptops, check that Voxxen is set to use the dedicated GPU in the NVIDIA Control Panel under "Manage 3D Settings → Program Settings." Also ensure no other GPU-intensive applications are running.
I hear crackling or distortion during playback.
This is usually a sample rate mismatch. Open Windows Sound settings and set both the input and output devices to 48000 Hz, 24-bit. Also check that no other applications are holding exclusive access to the audio device.
The level meter shows no signal.
Check that the correct device is selected in the INPUT DEVICE dropdown. Click the refresh button (↺) next to the dropdown if you plugged in the mic after launching Voxxen. Also check that Windows has granted microphone permission to the app.
Windows SmartScreen warns about the installer.
Click More info then Run anyway. This warning appears because the software is newly published. It is safe to proceed.

Contact Support

If you have a question not covered here, we are happy to help.

Email: [email protected]

We aim to respond within one business day.

Include in your support email: your license key or trial start date, a description of the issue, your GPU model, and any error message text you see. This helps us resolve things on the first reply.