Voiceover: from words to a finished mp3

1Where to find “Voiceover”

1. Open Studio

   The address is gibrid.ai/provideo. Sign in to your account.

2. Choose Audio → Voiceover

   In the top menu hover over Audio and choose Voiceover (caption “Text to speech with ElevenLabs voices”). The direct address of the screen is gibrid.ai/provideo/voiceover.

3. Take a look around

   At the top there is the title “Voiceover”, a Text / Dialogue switch and the My recordings button. Below are the text field, the “Voice” block, the “Quality” block and a bar pinned to the bottom with the price and the voice-over button.

2Voice a text in five steps

1. Write or paste the text

   Type into the field “Write or paste the text to voice…”. Under it are the buttons Insert (from the clipboard), Clear (right after clearing there is Restore), Emotions and the “Pronunciation language” choice. On the right is a counter: characters used out of the limit, the approximate duration and the price in ⚡. While the field is empty, an example “Welcome! Today we’ll show…” is offered.

2. Choose a voice

   The “Voice” block shows a card with the selected voice. By default it is the free “Бая”. Click Change or All voices to open the list. Below is a strip “Recent and favorites” for quick switching.

3. Choose the quality

   The “Quality” block has three cards: Fast, Natural, Maximum. Each shows a short hint and the price for your text. With free voices the cards are locked and say “free”.

4. Click “Voice over”

   At the bottom left you see the price in ⚡ and the number of characters, and under it the balance left. The Voice over button starts the synthesis (the quick way is Ctrl/⌘ + Enter). A seconds counter runs and Cancel appears next to the button.

5. Listen to the result

   A result card appears under the text: a ▶ button, a waveform and playback speed 0.75× / 1× / 1.25× / 1.5×. Buttons: Download mp3, In video, Another voice. The space bar plays and pauses the latest result when the cursor is not in a field.

3Voices: “Ready-made”, “Library”, “My clones”

The All voices button (or Change) opens the “Voices” panel. At the top are a search “Search: name, description, role” and a “Voice language” list; under them are three tabs:

• Ready-made — 47 voices. 12 paid ElevenLabs voices (for example “Брайан”, “Ксавьер”, “Джеймс”, “Марк”, “Белла”, “Ева”, “Моника”) and 35 free ones: Russian Silero (including “Бая”, “Ксения”, “Айдар”) and Piper, plus Kokoro in English, Spanish, French, Italian, Portuguese, Hindi and Chinese. Tab filters: “Free”, “Paid”, “Female”, “Male”.
• Library — ElevenLabs community voices, about 20 thousand, with filters; a click on a voice adds it for you. Details are on the Voice library page.
• My clones — your voice clones and the Create a voice clone button. Cloning needs the Pro plan or higher; details are on the Voice clone page.

Every voice has ▶ (listen), a “Favourite” star and a “free” or “paid” label. Recent and favourite voices are remembered in your browser, so on another device the list will be empty.

4Quality: “Fast”, “Natural”, “Maximum”

The quality level picks the ElevenLabs engine for you. It works for all paid voices: ready-made, from the library and your own clones.

Want a specific engine? Open Fine-tuning and in the “Engine” list choose one of four instead of “By quality”: “ElevenLabs v4”, “ElevenLabs v4 Turbo”, “ElevenLabs Flash v2.5” or “ElevenLabs Multilingual v2 (напрямую)”. Flash v2.5 costs the same as “Fast” (⚡4 per 1,000 characters), is the quickest and handles long texts, but does not understand emotion tags.

5Emotions and pauses

The Emotions button inserts a tag in square brackets where the cursor is, or over the selected fragment. There are 14 tags: Pause, Whisper, Laugh, Sigh, Excited, Joy, Sadness, Angry, Curious, Sarcastic, Nervous, Warm, Serious, Clear throat. In the text they look like [whispers], [laughs], [pause] and so on.

• Tags only work with “Fast” (v4 Turbo) or “Maximum” (v4) quality and a paid voice. If you press Emotions in another mode, the screen offers to Enable the right quality.
• With “Natural” or a free voice the server removes such tags so they are not read aloud. A line under the text warns you and has a Remove tags button.
• The v4 engines do not read the <break> pause tag: make pauses with commas, ellipses or the [pause] tag.
• Only “Maximum” (v4) understands pronunciation written in slashes as IPA, for example /ˈvɔɪs/.

The pronunciation language is set by the list next to the Emotions button: “Auto” (the language is detected from the text) or one of 13 languages — “Русский”, English, Español, Deutsch, Français, Italiano, Português, Türkçe, العربية, हिन्दी, 日本語, 한국어, 中文. If your language is not in the list, leave “Auto” and choose “Natural”: Gibrid does not send a language code to that engine.

6Fine-tuning

The Fine-tuning link in the “Quality” block opens the engine choice and four sliders. A dot appears next to it if you changed something; the values are remembered and Reset returns the defaults.

Free voices have no fine-tuning: the screen says “Free voices have no fine-tuning”.

7“Dialogue” mode

1. Switch to “Dialogue”

   At the top click Dialogue. Two lines appear with a male and a female paid voice.

2. Choose a voice for each line

   Click the voice name in the line header. A dialogue accepts any paid voice: ready-made, from the library and your clones. A free voice cannot be used: the screen says “Each line needs a paid voice”.

3. Write the lines

   The Line button adds one (up to 12 in total), the Move up / Move down arrows or dragging change the order, and Delete line removes one. You can write emotion tags in the lines, for example [laughs] or [whispers].

4. Click “Voice the dialogue”

   A dialogue always runs on the ElevenLabs v4 engine, and the bar shows “Dialogue · Maximum”. The whole dialogue is one entry in “My recordings”.

8Compare voices on your own text

1. Open “All voices”

   In the panel header click Compare.

2. Mark the voices

   Click two to three rows (“Select two to 3 voices”). A fourth cannot be added.

3. Click “Play in a row”

   The voices play one after another on the first ~70 characters of your text. Stop interrupts. The price of the whole comparison is shown at the bottom.

Free voices cost 0 in a comparison. For paid ready-made voices and clones the price is on the button (about ⚡1 per voice). A “Library” voice you have not added yet plays an ElevenLabs sample — that phrase is not yours and it is marked “sample”; it is free and the voice is not added to your list.

9“My recordings”: rename, download, go back to the text

Every voiceover is saved automatically (the caption “Saved to “My recordings””). The list opens with the My recordings button in the header; it shows only voiceovers — music lives elsewhere.

• Search “Search by name, voice, text” and the order Newest first / Oldest first.
• Click a recording’s name to rename it (Enter saves, Esc cancels). The server keeps up to 120 characters.
• Open text returns the text and voice of the recording to the editor, Use in video sends the track to Studio, Download gives the mp3.
• Deleting: click Delete and confirm with one more click (“Delete?”). The recording is removed for good.

10A voice track in video

The In video button on a result (or Use in video on a recording) puts the voiceover into the video creation bar in Studio: you get the message “Voiceover added to composer” and the main Studio screen opens.

Not every video model has a field for the track — only those that accept sound. The label depends on the model: “Voice track (optional)” — a ready-made voiceover drives lip sync and motion; “Voice sample for cloning” — the model takes the timbre and speaks the prompt. If the chosen model does not accept sound, there will be no field: change the model.

11What it costs and the limits

If you lack ⚡, the voice-over button stops with the hint “Not enough ⚡ — top up your balance” and a Top up link. More about balance and plans is on the Billing and plans page.

12What is not here

• There is no “Improve” button (AI director that places pauses and intonation) on this screen: you set pauses and emotions yourself with tags and punctuation.
• Free voices have no quality choice, no emotions and no fine-tuning, and you cannot pick a paid engine with them.
• You cannot clone someone else’s voice without consent; cloning needs the Pro plan or higher — see Voice clone.
• Professional voice cloning is not available yet: the fast clone is what works.

Related