All articles →
Speech synthesis

Speech synthesis: what it is and which tools to use in 2026

Speech synthesis turns written text into a voice: software reads text aloud and can save the audio to a file. Navigation apps, screen readers and video creators all use it. For expressive Russian narration with happiness, whispers, laughter and pauses, Dictator (dictator.my) offers Premium Voice. You control delivery with tags in the text, including Russian-language tags, while ordinary Microsoft voices read without a character charge.

BOZEX, dictator developer · Published

A woman wearing headphones listens to text read aloud beside a laptop and annotated notebook

Read in Russian →

We develop Dictator. The tag guide comes from the app's source code. The competing services below describe documentation and price checks on 14 October 2026, not a new November comparison. Dictator's billing information was updated from its pricing specification on 16 November 2026.

Try text to speech online with the free Microsoft voices, or download Dictator to use the desktop tag guide and Premium Voice.

How speech synthesis expresses emotion: a tag example

An expressive neural voice can follow instructions embedded in the text. Dictator's guide displays tags in the interface language. In the original Russian example, the voice follows the bracketed instruction rather than reading it aloud. An English rendering of the example is:

[happy] Hello! Today is a wonderful day. [laughing] Ha-ha, this is great! [whispering] Just don't tell anyone. [sighing] Oh well, time to get back to work.
Listen to the original Russian example: the same recording played by Listen in Dictator's tag guide · MP3

Put an emotion at the start of a sentence; it applies through the sentence. Put a sound where you want it to occur, and reinforce it with text such as “Ha-ha” or “Oh.” If a preset is not enough, describe the delivery in your own words, such as “[tired, almost whispering].” The linked audio is the original Russian recording, not newly generated English audio.

Speech-synthesis tools compared

For expressive narration, ask whether you can specify emotion and sounds in the text, and whether the model understands the language of those instructions. Competitor terms are the October 2026 snapshot; Dictator's billing is current to the repository check on 16 November.

ToolEmotion, sounds and pauses in the textFree accessAvailability from Russia in the reviewBest suited to
Dictator Premium VoiceThe app's guide includes 73 bracketed emotion, delivery, sound, pause and background tags; Russian tags and free-form instructions supportedMicrosoft voices have no character charge but do not follow these tags. A Free account includes 1,000 shared tokens, about three minutes if spent only on Premium VoiceYes; ruble and USD plan pricing availableExpressive narration, scenes and lessons
Speakmi website and Telegram botRussian tags for happiness, whispers and pauses; no full list on the reviewed siteThirty seconds after registration; then RUB 10–13 per minuteYes, ruble billingShort narration from a phone or bot
ElevenLabsEnglish audio tags such as [laughs], [whispers] and [sighs]; the reviewed documentation did not specify their behavior inside Russian textTen thousand credits a month, approximately ten thousand characters, without a commercial licenseThe reviewed help blocked access from RussiaExpressive English narration
Google Gemini-TTSEnglish tags in the preview documentation, such as [sigh], [laughing] and [whispering], plus delivery instructionsNo free Gemini-TTS allowance in the reviewed pricingNot specified on the reviewed page; billing requiredApp developers
Yandex SpeechKitNo bracketed emotion or sound tags in the reviewed workflow; some voices offer a style setting, marina supports whispering; text markup controls pauses and stressNo new-customer grant in the reviewed termsYes, through Yandex CloudRussian voice bots and applications
SaluteSpeech, SberSSML modes including whisper, sad, annoyed and happy for Russian text; no sound-event tags in the reviewed guideNo allowance for new customers in the reviewed termsThe service had stopped connecting new clients from 15 July 2026Previously connected companies
Microsoft AzureSSML styles; the reviewed Russian voices Svetlana, Dmitry and Darya did not list stylesFive hundred thousand characters a month in the reviewed tierNot specified on the reviewed pageApp developers
OpenAI gpt-4o-mini-tts, reviewed documentationDelivery controlled with a separate natural-language instruction rather than the bracketed-tag workflow used hereNot specified in the reviewed pageRussia absent from the supported-country listSpeech generation in applications
A phone's built-in speech synthesisVoice, rate and pitch rather than these embedded performance tagsIncluded with the deviceYesReading a screen aloud

In the October comparison, Dictator Premium Voice and Speakmi both documented Russian instructions in the text. Dictator also supplied a complete desktop guide explaining tag placement. Try text to speech online or download Dictator to try the app workflow.

Choose a speech-synthesis tool for the task

How speech synthesis works

A speech-synthesis system has three broad jobs. First it prepares text: expands numbers, dates and abbreviations, and decides pronunciation and stress. Then it determines delivery, including pauses and the rise or fall of the voice in a question or statement. Finally, the model generates audio.

System voices in Android, iPhone, Windows and macOS are included with the device, but do not use the expressive instruction workflow described here. Cloud neural voices can produce expressive speech, but send text to a server and usually meter usage by characters, credits, tokens or generated audio. Local voices are downloaded once and then read offline; Dictator includes local-voice options in its “On this computer” section.

There are three common ways to control delivery: developer-oriented SSML, where a pause may be written as <break time="1s"/>; bracketed tags like those above; or a separate instruction such as “Read in a whisper, with pauses.” The accepted syntax depends on the model.

Dictator's tag guide: emotion, delivery, sounds, pauses and background

The “Tags: delivery and sounds” guide groups 73 bracketed tags into six categories, adds emotion-strength examples, and describes seven paired tags for Premium 3 Drama. Below are English renderings of the categories in the original guide; the app supplies tags in its interface language.

Two tags can be placed together, such as “[sad][whispering].” Contradictory combinations, such as happiness and sadness at the same point, are less predictable; describe complex delivery in a single instruction instead. The voice can also follow instructions in other languages. Dictator's interface supports nineteen languages and displays the guide accordingly.

Dictator's tag guide: an example with a Listen button and emotion tags copied by clicking
Dictator's tag guide: an example with a Listen button and emotion tags copied by clicking

Generate expressive speech in Dictator in three steps

  1. Download Dictator and sign in to a free account to use cloud Premium Voice. Open Text to Speech and choose Premium Voice as the engine.
  2. Paste your text and open “Tags: delivery and sounds” above the text box. Click a tag to copy it, then paste it into the text where it should apply.
  3. Choose a voice and generate the audio. You can listen and save the result in MP3, WAV or OGG.
Dictator Text to Speech: tag guide above the text, with Premium Voice, model selection and the account balance on the right
Dictator Text to Speech: tag guide above the text, with Premium Voice, model selection and the account balance on the right

Premium Voice uses the shared account token balance. A Free account includes 1,000 tokens, approximately three minutes when the whole balance is spent on that voice. Basic includes 20,000 monthly tokens for $5.99, Pro 46,000 for $12.99, and Premium 100,000 for $29.99. Those tokens are shared with other cloud features, not separate guaranteed voice-minute buckets. Paid accounts can buy token top-ups, valid for 180 days; new minute packs are no longer sold. Previously purchased minute packs remain usable until their expiry. Both voice models use this billing path. Microsoft and suitable offline voices remain free and do not understand these Premium Voice performance tags. See current pricing.

Premium 2 Pro versus Premium 3 Drama

Premium Voice offers two model choices. Premium 2 Pro is the stable default and supports the emotion and sound instructions described by the guide.

Premium 3 Drama is labeled as a preview in the app. It is designed for more descriptive delivery instructions: emotion, intention, pace and whom you are speaking to, such as “[warmly, as to an old friend] Hello, it's been a long time!” The guide describes its pause instruction as more audible. It also supports paired tags affecting selected words. For example, in “<stress>You</stress> did it. <whisper>Just don't tell anyone.</whisper>,” the selected word is stressed and the second sentence whispered.

The seven paired tag names are whisper, emphasis, stress, soft, forceful, slow and fast. Their names are always written in English, even in another interface language. Drama's output can vary between generations, so listen to important lines before using them.

How other tools control delivery

The ElevenLabs documentation examined for the October review described bracketed tags including [whispers], [laughs], [sighs], [sarcastic] and sounds such as [applause], and using uppercase text for emphasis. Its models supported Russian, but the reviewed documentation did not explain the behavior of English tags inside Russian passages. The checked free plan provided ten thousand credits a month and Starter cost $6 a month. Its help, dated 24 September 2026 in the original review, stated that access from Russia was unavailable. Consult current documentation and plans before relying on those historical terms.

Yandex SpeechKit is aimed at developers. In the reviewed markup, a pause was written as sil<[500]>, with milliseconds inside the brackets, and a plus sign before a vowel marked stress. Emotion was a voice setting rather than this article's bracketed-tag syntax. The voice marina offered whispering, while jane offered cheerful and irritated styles.

What the speech-synthesis setting on a phone means

On a phone, speech synthesis usually means the built-in function that reads text aloud. Android exposes the engine, language, rate and pitch through its accessibility or text-to-speech settings; exact menu names depend on the device. You can choose Google's engine, the manufacturer's engine or another installed app. Download a missing language through the voice-data installation option.

On iPhone, spoken-content accessibility features can read selected text or the screen. In the workflow described by the original review, Speak Selection read a highlighted passage, and Speak Screen read the screen after a two-finger downward swipe from the top. These system settings control voices and reading speed rather than the expressive laughter, whispers and emotion instructions in this article.

Where speech synthesis can make mistakes

The input text itself is often the problem. A model infers stress in words with identical spelling from context; Russian words such as the two pronunciations of “zamok,” meaning a lock or a castle, illustrate the difficulty. Abbreviations may sound odd, while their fully written forms read more naturally. A foreign name in Latin characters may be unpredictable for a Russian voice, whereas an appropriate phonetic Russian spelling can make the intended pronunciation clearer.

Punctuation is the first way to shape delivery: commas and periods suggest pauses, while a question mark changes the end of a phrase. A tag adds a performance that punctuation alone cannot express, but it does not replace a well-punctuated paragraph.

Frequently asked questions about speech synthesis

What is speech synthesis in simple terms?

It is a computer or phone reading written text aloud. The system handles numbers, abbreviations, pronunciation and pauses, then generates audio that you can listen to or save, for example as MP3.

Which speech-synthesis app is useful for expressive Russian narration?

Dictator Premium Voice provides a guide to emotion, sound and pause instructions, including Russian-language tags, and runs through the desktop app with shared-token billing. For a Russian voice bot, Yandex SpeechKit is a developer-oriented option with its own pronunciation and pause markup.

Do emotion tags work in Russian?

Dictator's guide provides Russian instructions for whispering, laughter, sighing and other effects. The model follows them instead of reading the bracketed instructions aloud. The guide covers both Premium Voice models, with paired tags reserved for Premium 3 Drama. The competitor comparison describes the limits of the documentation we checked in October, not proof that other models cannot do it.

Can I use speech synthesis for free?

Yes. Dictator's standard Microsoft and suitable offline voices are free without a character charge. The reviewed catalogue contained 322 Microsoft voices across 75 languages and four local Russian voices. A Free account's 1,000 tokens can also be tried on Premium Voice; they represent about three minutes if the entire balance is spent there. Paid plans and top-ups use the same shared token balance rather than the retired minute packs.

How do I stop a phone from reading everything aloud?

A screen reader such as TalkBack is often responsible and is controlled separately from choosing a text-to-speech engine. The Google help used in the original review explained that a configured accessibility shortcut could toggle TalkBack by holding both volume buttons, or that Google Assistant could respond to a request to turn it off. The available shortcut depends on the phone's settings.

Listen to the difference: try text to speech online, then download Dictator and experiment with the tag example.

Related: free text to speech without a character limit, different voices and character dialogue, English text to speech, and free voice cloning. Competing-service documentation was checked on 14 October 2026. Reference pages used in that review include ElevenLabs, Google Gemini-TTS, SpeechKit markup and Azure speech markup. Dictator's billing and model guide were checked against its source on 16 November 2026.

Who wrote this

BOZEX develops Dictator (dictator.my), an app for text to speech, voice typing, transcription and translation. The Premium Voice tag guide comes from the app's source. Other services were checked against their documentation and pricing on 14 October 2026; Dictator's billing and model details were checked in its repository on 16 November 2026.