Voice Changer Studio User Guide

Quick start

Once setup is done, every project follows the same eight steps.

1. Open the app and enter your app password if asked.
2. Drag your video or audio file onto Your video or audio (or click it to browse).
3. In Pick a voice, filter by gender, accent and age, press the play button to hear samples, then click Use.
4. Press Write out the speech. Every spoken line appears in the script as timed, editable text.
5. Edit the script: type over lines, Remove lines, add new ones with + Line, and use Hear line to check wording.
6. Choose a voice mode, then press Preview 20 seconds to hear the new voice on your own clip.
7. Press Make new voice track. Watch it and switch between Original and New voice to compare.
8. In Download, pick a format and press Download. The file lands in your Downloads folder.

One-time setup

Setup takes about 10 minutes and is done once. The app lives in the GitHub repo MyKyah70/Voices and runs on Cloudflare Pages; the voices come from your ElevenLabs account.

1. Get an ElevenLabs API key

1. Sign in at elevenlabs.io. A paid plan (Starter or higher) is required for Voice Library voices and voice swapping.
2. Go to elevenlabs.io/app/developers/api-keys (or the left sidebar: Developers > API Keys).
3. Click Create API Key, name it Voices, and turn on: Text to Speech, Speech to Speech, Speech to Text, Voices (Write), User (Read).
4. Copy the key right away. It is shown only once; if lost, create a new one.

2. Connect Cloudflare to GitHub

1. In Cloudflare: Workers & Pages > Create > Pages > Connect to Git, then pick MyKyah70/Voices.
2. Framework preset: None. Build command: leave blank. Build output directory: public.
3. Click Save and Deploy.

3. Add the secrets

In the Pages project, open Settings > Variables and Secrets and add both as type Secret:

4. Redeploy

Go to Deployments, open the menu on the latest one and choose Retry deployment. New secrets only take effect after a deploy. Every push to GitHub also redeploys automatically.

The screen

The screen has four numbered panels. On a computer, panels 1 and 4 stay pinned on the left while you scroll the script on the right, so the video is always in view.

The top bar shows your remaining ElevenLabs credits, a Password button if a password is set, and the User Guide button that opened this page.

Keyboard: Space plays and pauses the video whenever you are not typing in a box.

Opening a file

Drag any video or audio file onto panel 1, or click it to browse. To switch files later, press Open another file.

What happens behind the scenes: formats your browser plays natively open in a few seconds. Anything else goes through the built-in converter, which downloads once (about 30 MB), then makes a viewing copy and pulls out the sound track. The file info line under the player says converted for viewing when this happened. Your original file is never changed.

Large files: files over about 700 MB use the converter to read the sound, which can take a minute or two. Very long or 4K files work, but plan for the download step to take longer (see Downloading).

Coming back to a file: if you reopen the same file in the same browser, your saved script edits are restored automatically.

Choosing a voice

Pick the voice before you preview or build; you can change it at any time and rebuild.

• Voice Library tab: thousands of real-person voice models shared on ElevenLabs, sorted by what is trending.
• My Voices tab: voices already in your ElevenLabs account, including any you used before or cloned yourself.
• Filters: Gender (male, female), Accent (American, British, Australian, Irish, Scottish, Southern, Indian and more) and Age (young, middle aged, older).
• Search: type a description such as narrator, deep, warm, gravelly or cowboy.
• Listen: press the round play button on any voice to hear its sample. Press it again to stop.
• Select: click Use. The chosen voice is outlined and named under the list.
• Load more appears at the bottom when more matches are available.

Keep the same person's voice (edit their words)

To change what someone says without changing who is speaking, press Use the original speaker's voice (below the voice list).

1. Open the video and press Write out the speech first; the app then copies the voice from the actual speech (up to 2 minutes, at least 10 seconds needed).
2. Press Use the original speaker's voice and confirm you are the speaker, own the recording, or have their permission.
3. Edit, add or remove lines, then press Make new voice track. Only your edited lines are spoken in the copied voice; everything else stays exactly as recorded.

The copy is saved in your ElevenLabs account (under My Voices) and reused next time you open the same file. It uses one of your plan's custom voice slots.

The first time you use a Library voice, the app adds it to your ElevenLabs account automatically; after that it also appears under My Voices. Your plan limits how many custom voices you can hold, so remove unused ones in ElevenLabs if you hit that limit.

Editing the script

Press Write out the speech and the app writes out everything said in the file as a list of timed lines. Long files are processed in parts; a 20 minute video takes about a minute. Press Write out again to start over (it asks first if you have edits).

Each line has its start time on the left, the words in the middle and its actions on the right.

Status tags: each line shows Original, Edited (orange edge), New line (green edge) or Removed (crossed out). Edited lines show what they used to say underneath. After a build, a red tag such as 1.2s too long means the new wording does not fit before the next line; shorten it or move the next line later.

Follow along: with Follow along while playing ticked, the script scrolls to keep the current line in view and highlights it in blue.

Saving your work

• Edits save automatically in this browser, per file. Reopening the same file restores them.
• Save script downloads your script as a .json file; Open script loads one back, for example on another computer.
• Subtitles (SRT) downloads captions of the current script for YouTube or any video editor.

Voice modes and fine tuning

Choose Keep original delivery for lip sync and natural emotion; choose Fresh read, full accent when the new accent matters most.

Fine tune the voice (click to expand in the app)

• Stability: lower gives more emotion and variety; higher gives a steadier, calmer read. Start at 0.50.
• Voice match: how closely the result sticks to the chosen voice. Start at 0.80.
• Expressiveness: adds style and drama. Leave at 0 for natural speech; raise slowly for narration.
• Clean up background noise: removes hiss and room noise from the original before swapping. Turn it off to keep room sound.

Changing the voice, mode or any of these settings means the next build regenerates the affected audio.

Previewing and making the new voice track

Preview a few spots first, then build the whole track once you like the voice.

Preview 20 seconds

1. Move the player to a spot with speech and pause.
2. Press Preview 20 seconds. The app builds the new voice for the next 20 seconds only and starts playing it.
3. Try a different voice, mode or setting and preview again. Each preview costs about 300 credits.

Make new voice track

1. Press Make new voice track. Progress shows in the bar and status line, line by line.
2. When it finishes, the player switches to New voice and plays from the start. Captions show each line.
3. Use Original and New voice under the player to compare at any moment.
4. Make more script edits, then press Update new voice track. Only changed lines are regenerated; untouched parts are reused, so updates are fast and cheap.

How edits are blended: in Keep original delivery mode, the original wording of each edited or removed line is faded out, and the new line is placed at its start time. If a new line is longer than the gap, it is spoken slightly faster (up to 20%) to fit; anything still too long is flagged in red.

Downloading

Download unlocks after a full Make new voice track (a preview is not enough). Pick a format in panel 4 and press Download.

Video quality: for files under 1.5 GB, the picture is copied as-is with the new sound track, so quality is untouched and the download takes seconds.

Audio-only files can only be downloaded in audio formats.

Credits and costs

The biggest cost is a full build in Keep original delivery mode: roughly 1,000 credits per minute of video, charged once per voice and settings. Figures are approximate; ElevenLabs Usage analytics shows exact use.

Example: a 20 minute video uses about 20,000 credits for the full build, plus a few hundred per batch of edits. The Starter plan's 40,000 monthly credits cover about two such videos.

Tips for the most natural result

• Match the speaker. A voice close in gender and age to the original sounds most believable in Keep original delivery mode.
• Keep edits about the same length. A new line with roughly the same number of words as the old one fits the gap and the lip movement best.
• Write the way people talk. Use contractions and natural punctuation; commas and periods shape the pauses.
• Spell out tricky words the way they sound and check them with Hear line.
• Use punctuation for emotion. An exclamation mark or question mark changes the read; an ellipsis adds a hesitation.
• Turn stability down (0.30 to 0.45) for lively, emotional scenes; up (0.60 to 0.75) for calm narration.
• Clean source audio helps. Wind, music or overlapping voices in the original reduce quality; leave Clean up background noise on unless you need the room sound.
• Background music is replaced. The new track contains voice only. Add music back afterward in a video editor if needed.
• Only voice content you own or have permission to change, and do not use a voice to imitate a real person without their consent.

Troubleshooting

Most problems are fixed by a redeploy in Cloudflare or a hard refresh (Ctrl + F5) in the browser.

If something else goes wrong, take a screenshot of the status message and the browser window and send it to whoever maintains the app.