A clear definition and a practical example for every term, with a link to the full guide.
A recording played when customers call outside working hours or on holidays, giving opening times and an alternative such as the website, WhatsApp or an emergency line.
A tool that turns a written idea into an animated cartoon episode: it writes the script, casts the roles, draws the shots and adds voices, sound effects, music and captions, with no drawing or editing skills needed.
A tool that turns text into speech in a cartoon character's voice, such as a hero, a villain or a narrator, instead of recording a voice actor. Some generate single lines; others are part of a cartoon studio that knows the script and who is speaking.
A version of a Japanese anime series voiced in another language instead of the Japanese cast. Arab audiences know shows like Detective Conan, Dragon Ball and One Piece through their Arabic dubs.
A secret token that identifies your account when calling an API and against which usage is billed. Keep it on your server, never in app or website code users can see.
The spoken form of Arabic in a particular country or region, such as Saudi, Emirati, Kuwaiti, Egyptian or Levantine. Dialects differ in pronunciation, vocabulary and some grammar, and matching your audience's dialect makes a voice sound local.
Arabic written in Latin letters, with digits standing in for sounds English lacks: 3 for ain, 7 for haa, 5 for khaa and 2 for hamza. It is common in chat, but TTS engines read Arabic script far more reliably.
The ratio of a video's width to its height. 9:16 is vertical for TikTok, Reels, Shorts and Snapchat; 16:9 is horizontal for YouTube and screens; 1:1 or 4:5 suit feed posts.
AI-generated audio that puts words in a real person's mouth they never said, in order to deceive, defraud or defame. What separates it from legitimate cloning is consent and transparency.
An inaudible signal embedded in an audio file that special tools can detect, to show the audio was generated or to trace where it came from.
A simpler form of IVR that answers calls and transfers them to the right department or extension based on the key the caller presses, without looking up data or completing transactions.
Supplementary footage shown over speech or voiceover to illustrate it or keep the pace up, such as product, location and detail shots, as opposed to the main shot of the speaker.
How many bits describe each audio sample, which sets the range between the quietest and loudest sound before distortion. 16-bit is the CD standard; 24-bit is common for recording and editing because it leaves more headroom.
How much data each second of audio uses, measured in kilobits per second (kbps). In compressed formats like MP3, a higher bitrate means better quality and bigger files.
The consistent voice a brand uses across ads, phone messages and videos so audiences recognise it by ear. It can be a library voice, a designed voice or a clone of a consenting speaker.
Text drawn into the video picture itself, so it is always visible and cannot be switched off or edited without re-exporting. Ideal for TikTok and Reels, where many people watch with the sound off.
The line that tells the listener exactly what to do next: order, book, call, tap the link. It usually closes the ad and works best as a single clear action.
Replacing cartoon characters' voices with voices in another language or dialect, matching emotion and timing. Arabic cartoon dubs of the 80s and 90s were mostly in simplified Classical Arabic, which is why they became part of a generation's memories.
A ready-made starting point in an AI cartoon maker: a show's world with its cast, character voices and art, so you only write a new episode idea. It differs from a general style, where the characters are designed for your story.
A reference drawing of a character from several angles, with fixed clothes and colours, used for every shot and episode so the character stays on model. Ready-made series casts come with their sheets for free.
The voice that sets a cartoon character apart: its pitch, pace, energy, accent and emotional tone. In AI cartoons each character keeps one consistent voice across every episode.
Distortion that happens when audio exceeds the highest level a file can represent (0 dBFS), so the tops of the waveform are cut off and the sound turns harsh. It cannot be fully repaired, so lower the level before recording or exporting.
Captions stored in a separate file such as SRT or VTT that viewers can switch on or off in the player. They let you add several languages to one video and edit text without re-exporting.
Moving between two languages within the same sentence, as many young people and professionals in the Gulf do. It needs a voice that pronounces the English words naturally without breaking the Arabic rhythm.
Permission to use generated audio in paid or business work, such as ads, monetised videos and client projects. They are set by the platform's terms and your plan, so check them before publishing commercial work.
Vocla's usage units, deducted when you generate voice, video and other output depending on the feature and length. The Free plan includes 500 credits a month, Starter 5,000 and Professional 80,000.
The spoken Arabic of Morocco, Algeria and Tunisia, with strong Amazigh and French influence. Many Gulf and Levantine listeners struggle to follow it quickly, so campaigns aimed at both regions usually need separate versions.
Dual-tone multi-frequency: the tones a phone sends when keys are pressed, with each key producing two frequencies at once. It is how an IVR knows which option the caller picked.
Replacing the original speech in a video with speech in another language or dialect, timed to the picture as closely as possible. Unlike subtitles, the viewer hears the new language instead of reading it.
Automatically lowering music whenever speech starts and bringing it back when speech stops, so the voiceover stays clear without riding the fader line by line.
The dialect of Egypt, specifically Cairo, and one of the most widely understood in the Arab world thanks to film and TV. It pronounces jeem as a hard g and qaf usually as a glottal stop; common words are ayez (I want), dilwa'ti (now) and keda (like this).
Video content where the creator never appears on camera, relying on voiceover, footage or generated scenes and on-screen captions. Common for explainer, story and review channels.
How many images are shown per second (fps). 24 fps feels cinematic, 25 and 30 are standard for TV and social, and 60 suits fast motion.
An ITU-T audio codec for telephone calls: 8 kHz sampling, 8-bit companded samples and 64 kbit/s. It comes in two variants, μ-law and A-law, and many phone systems require it for prompt files.
The dialect of western Saudi Arabia, in Jeddah, Mecca and Medina, distinct from Najdi in sound and vocabulary. Typical words include eish (what), abgha (I want) and dahheen (now).
The first one to three seconds of a short video, whose only job is to stop the scroll. On TikTok and Reels it is usually a question, a surprise or a clear promise.
Generating a short moving clip from a still image, such as a camera move around a product or added motion in the scene. Useful when you have good product photos but no footage.
Making a dubbed line start and end at roughly the same moments as the original. Arabic translations can run longer than English, so the wording is tightened to fit rather than speeding up the voice.
A phone system that greets callers and offers a menu they navigate with keypad presses or speech, then routes them to the right team or plays recorded information.
Each recorded audio clip a caller hears in a phone system: the greeting, the menu, error, hold and closing messages. They are recorded or generated as separate files and uploaded to the phone system.
An umbrella name for the dialects of the Gulf states: Kuwait, the UAE, Qatar, Bahrain, Oman and eastern Saudi Arabia. They share many words such as wayed (a lot) and shlon (how), but each country has a flavour local listeners spot immediately.
The time between sending a request and receiving the first piece of audio. It barely matters for pre-made files such as ads, but voice assistants and live calls need it very low to avoid awkward silence.
The dialects of Syria, Lebanon, Jordan and Palestine. They differ from each other but share words like shu (what), biddi (I want), halla' (now) and ktir (very). They come across as soft and friendly in ads and social content.
Matching dubbed speech to the on-screen speaker's mouth movements. Some tools only adjust timing and wording; others alter the mouth in the picture with AI. It matters most in close-ups.
Adapting content fully for a specific market: language, dialect, currency, dates, examples, images and cultural references, not just the translation.
Loudness Units relative to Full Scale: a measure of how loud a whole file sounds to the ear, not just its highest peak. European broadcast targets −23 LUFS, streaming platforms normalise to around −14 LUFS and podcasts often aim near −16.
The standard modern form of Arabic used in news, books, official documents and education across the Arab world. Everyone understands it, but it sounds more formal than the spoken dialects in ads and social content.
Mono is a single channel heard identically in both ears; stereo has separate left and right channels. Voiceover is normally recorded and generated in mono, while music is usually stereo.
A compressed (lossy) audio format that discards sound the ear is unlikely to notice, making files much smaller for a small loss in quality. Great for web and social, not for repeated editing.
Attaching several audio tracks in different languages to one video, so viewers pick their language in the player instead of you publishing a separate upload per language. Platforms such as YouTube offer this to many channels.
A track placed under a voiceover to set mood and pace without competing with the words. It must be licensed for your use and sit clearly below the voice.
The dialect of central Saudi Arabia, around Riyadh and Qassim, and what most people hear as "Saudi". It often pronounces qaf as a hard g and uses words like wesh (what), abi (I want) and alheen (now).
The voice that tells the story from outside the action: it opens the episode, links the scenes and closes with the moral or a "to be continued". It is usually warm, steady and slower than the characters' voices.
Traditional telephone audio sampled at 8 kHz, carrying roughly 300 to 3,400 Hz, which is why calls sound thin and muffled. HD Voice (wideband) calls use 16 kHz.
Text to speech built on neural networks trained on many hours of human speech, so intonation and pauses are generated rather than stitched together from recorded fragments. It is why modern AI voices sound far closer to a real narrator than older robotic systems.
The level of sound left in a recording when nobody is speaking: air conditioning, a laptop fan, traffic. Lower is cleaner; ACX audiobook specs, for example, require a noise floor no higher than −60 dB.
Processing that removes or reduces unwanted sound such as hum, hiss and room echo from a speech recording. Modern AI tools separate the voice from the background, but overdoing it makes speech sound metallic.
The recording callers hear while waiting to be answered, usually music mixed with short messages about services, offers and reassurance that their call matters.
A licence that allows generated audio to be used for non-commercial purposes only, such as testing, learning and private projects. At Vocla the Free plan is personal use; paid plans from Starter up include commercial use.
The smallest unit of sound that changes a word's meaning when swapped. Some TTS tools let you spell a word in phonemes to fix one the engine keeps mispronouncing.
Re-recording a line or section after the session to fix a mistake or a script change, matched in voice and tone to the rest. With a generated voice, a pickup is just regenerating one line instead of booking a new session.
Bursts of air from sounds like b, t, d and k that hit the microphone and come out as a low thump or pop. They are tamed with a pop filter or by speaking slightly off the mic's axis.
A saved list of how specific words should be said, such as brand names, people's names and foreign terms, which the engine applies automatically wherever the word appears.
The rise and fall, speed, rhythm and stress of speech. It is what makes the same words sound like a question, a statement or excitement, and weak prosody is the quickest giveaway of a low-quality synthetic voice.
The maximum number of requests you may send to an API in a given period. Exceed it and the server usually answers with HTTP 429, so you retry after a short wait.
A licence where you pay once or as part of a subscription, then use the material without per-play or per-broadcast fees. It does not mean free or copyright-free; the licence terms still apply.
The part of a video frame not covered by app buttons, the account name and the caption text. On TikTok and Reels parts of the bottom and side are covered, so captions and key text should stay clear of them.
How many times per second audio is measured when it is stored digitally, given in kilohertz (kHz). Traditional phone lines typically use 8 kHz, music uses 44.1 kHz and video usually uses 48 kHz.
Harsh, exaggerated s, sh and z sounds that stand out more in a recording than in normal speech. A de-esser or a change of mic angle tames it.
Using consistent audio elements to identify a brand: one narrator voice, a short jingle or sound logo, a greeting tone and a way of speaking, everywhere customers hear the brand.
A narration style that recalls the 90s dubbed cartoons on Arabic TV: a deep, warm voice in clear Classical Arabic, short epic lines, an opening that hooks kids and a closing with a moral or a cliffhanger.
Splitting a recording by speaker to answer who spoke when, so the transcript comes out as turns: Speaker 1, Speaker 2.
Technology that turns spoken audio into written text, also called automatic speech recognition. For Arabic its accuracy depends on dialect, recording quality, overlapping speakers and English mixed into speech.
A plain-text subtitle format where each caption is numbered and has a start and end time. Most video platforms and editors accept it; VTT is a close relative used by web players.
Speech Synthesis Markup Language, an XML-based W3C standard for controlling synthetic speech: pauses, emphasis, rate, pitch and how numbers or dates are read. Which tags work depends on the engine.
A video broken into ordered scenes, each with a visual description, its line of script and its duration. In AI video you review and approve it before rendering, which saves time and credits.
Sending audio in chunks as it is generated rather than waiting for the whole file, so playback starts almost immediately. Essential for live conversation, voice assistants and long texts.
Any voice generated by a computer rather than recorded directly from a person, whether it is a stock voice from a library, designed from a description or cloned from someone with their consent.
One attempt at reading a script or part of it, whether recorded by a narrator or generated by a voice tool. You usually make several takes and keep the best, or combine parts of them.
The episode's stills in order with their voiced lines, reviewed before production. In Vocla Cartoon Studio it is the last review point before anything is charged, and you can redraw any shot for free, up to 30 times per episode.
The vowel and pronunciation marks written above or below Arabic letters: fatha, damma, kasra, sukun, shadda and tanween. Most Arabic is written without them, so a text to speech engine has to guess; adding tashkeel to ambiguous words tells it exactly how to read them.
How many people can sign in to and work on the same plan, each with their own login. Vocla's Professional plan includes 3 seats.
The step where a speech engine turns numbers, dates, currencies and abbreviations into words before reading them. Arabic makes it harder because numbers agree in gender and case with the noun they count.
Generating a conversation between two or more voices from a single script split by speaker, with each speaker keeping their own voice and the turns flowing naturally.
Technology that reads written text aloud in a synthetic human voice and outputs an audio file. For Arabic, quality depends on how well the engine handles missing diacritics, numbers and the dialect you want.
Generating video with AI from a script or written description. Some tools produce visual scenes only; others build a full video with voiceover, scenes and on-screen captions.
Times attached to a transcript showing when each sentence or word was spoken in the recording. They are what turn a transcript into captions synced to the video.
Carrying an advertising message into another language by rewriting it to have the same effect in the new culture, rather than translating it literally. It is used for slogans, hooks, jokes and wordplay.
A way for your software to send text to a voice service programmatically and receive audio back, without using the web interface by hand. It suits generating thousands of messages or voice inside an app or phone system.
In human voiceover, a fee paid to the voice actor for using the recording in specific media, for a set period and territory, on top of the session fee. A year of TV across several countries costs more than an internal video.
The written description that guides a video model: subject, setting, lighting, camera movement and style. The more specific and visual it is, the closer the result is to what you imagined.
The number of pixels in a video frame. 720p is 1280×720, 1080p is 1920×1080 and 4K is 3840×2160; in vertical video the numbers flip, so 1080p becomes 1080×1920.
A system that converses with users by voice in real time, combining speech to text, a language model that understands and replies, and text to speech. It needs low latency and a solid grasp of the caller's dialect.
The stage where voices are assigned to characters and auditioned before recording or production. Good casting keeps voices apart in pitch and energy so each character stands out.
A tool that takes a recording of you and re-renders it in a different voice while keeping your performance: timing, pauses and emotion. Technically it is called speech to speech.
Creating a synthetic voice that sounds like a specific person from recordings of them, then using it to read new text. It is used for brand voices and by creators who want a digital copy of their own voice, and it requires the speaker's consent.
Clear, recorded permission from a person before their voice is cloned or used, stating what the voice may be used for and for how long. Without it you should not clone anyone's voice, even if their recordings are public.
Creating a brand-new synthetic voice from a written description, such as age, tone and character, without recordings of any real person. It suits brands that want a distinctive voice without cloning a human.
Notes describing how a script should be read: tone, pace, energy, which words to stress and who the audience is. They go at the top of the script or in brackets before the line they apply to.
The set of ready-made voices a platform offers, usually filterable by language, dialect, gender, age and tone, so you can pick one without recording or cloning anything.
Choosing a voice for the dubbed version that resembles the original speaker in gender, age, energy and character, or reproducing their own voice in another language with consent, so face and voice do not feel mismatched.
The recording a voice clone is built from. The cleaner it is, with no noise, music or echo, and the closer the delivery is to how you will use the voice, the more accurate the clone.
The voice of an unseen narrator who explains, tells or persuades over images or video. It is used in ads, training videos, documentaries and social content.
Text written to be heard rather than read: short sentences, one key idea per line, numbers and names spelled the way they are spoken, plus tone and pause notes in brackets.
A W3C caption format built for web video players. It resembles SRT but starts with a WEBVTT line, uses a full stop instead of a comma before milliseconds and supports more styling and positioning.
An audio file format that usually stores sound uncompressed, so it keeps full quality but is much larger than an MP3. It is the safer choice for editing, phone systems and anything you will process further.
A URL on your system that a service calls automatically when a long task finishes, instead of your system asking every few seconds whether it is done. Common for slow jobs such as dubbing and video generation.
The standard accuracy measure for speech to text: substituted plus deleted plus inserted words, divided by the number of words in the correct transcript. Lower is better; in Arabic, spelling variants of hamza and taa marbuta also affect the score.
A softened spoken Arabic that drops very local words and leans towards standard vocabulary while keeping a colloquial sound and rhythm, so a wider audience understands it. It is common in ads aimed at several Gulf countries.
A reading-speed measure used to estimate how long a voiceover will run from its script. English voiceover sits around 150 words per minute; Arabic counts fewer words because pronouns and prepositions attach to the word itself.
The two companding methods used in G.711 to store quiet sounds with more precision. μ-law is used in North America and Japan, A-law in Europe and most of the rest of the world; mixing them up makes prompts sound harsh.
Write your script, pick the dialect and voice, and download the file in minutes. Start free; paid plans from $1/month.