Speechify and Voicemaker: 12 neural networks for text-to-speech

ElevenLabs
The largest platform specializing in creating AI for speech synthesis. Their neural networks are used for narrating audiobooks, dubbing videos, and creating voice assistants. The site is only accessible via a foreign IP address. The voices are lifelike and sound like audiobook voice actors. Each reads as if with the right emotions, intonation, and pauses.

Murf
Offers a wide selection of AI voices with natural sound in over 20 languages for professional video and presentation narration. The service makes it convenient to compare results side by side—you can create blocks, select up to 10 voices at once, and listen to them one after another.

Speechify
Scanning and narration—can convert not only digital text but also text from paper documents into audio. Premium-class voices—offers natural voices, including celebrity imitations. Speed adjustment up to 900 words per minute—ideal for rapid information consumption.

CapCut
Built-in speech synthesis in the video editor—allows instant text narration directly within the project without needing to upload separate audio files. Variety of voices—includes several narrator options with different intonations and styles. Automatic subtitle creation—can generate subtitles based on text or audio files. Editing flexibility—allows adjusting tempo, pitch, and applying effects.

Voicemaker
The voice database is truly extensive—featuring several hundred voices, including both monolingual and multilingual narrators. Settings allow adding pauses, changing volume and speed, and imparting specific intonations. The result can be downloaded in MP3, WAV, OGG, AAC, or OPUS formats.

NaturalReader
Text can be narrated using a dozen different voices that vary by accent, emotion, age, and gender of the narrator. There is a mobile app that allows you to even read books using NaturalReader by showing pages through the camera.

Yandex SpeechKit
Yandex SpeechKit can narrate texts in Russian, Kazakh, Uzbek, English, German, and Hebrew. Eleven voices are available for selection, including male and female options. Settings allow adjusting speech speed and specifying style: neutral or friendly.

SteosVoice
A specialized voice-over service based on neural network models, accessible via a Telegram bot. SteosVoice offers high-quality speech synthesis with over 400 voices, including those of professional actors, cartoon characters, and video game characters. Settings allow adjusting voice speed and pitch, as well as file format.

Voicer
In the free version, you can narrate texts up to 1000 characters, and the voices maintain a neutral sound. However, you can adjust parameters such as bass, echo, speed, pitch, frequency, and volume to find the optimal sound for your tasks.

Freetts
A total of 29 Russian-language voices are available. They are quite diverse, and the text is narrated without errors and even with correct intonation, but the unnaturalness is strongly noticeable. There is a separate "Maxim-bot," whose voice you will definitely recognize from YouTube and TikTok videos.

Apihost
The service has seven versions of neural networks that differ in character limits and the number of voices. For example, V1 has 1000 characters and 17 voices, while V2 has 500 characters and 20 voices. The models differ in generation quality, emotion set, and processing speed.

Zvukogram
Choice of narrators, supports male, female, and child voices of different tones. Speed and intonation adjustment—allows setting speech tempo and emotional tone. Business voice-over generation—suitable for creating voice announcements, autoresponders, and video clips. Working with long texts—supports narrating large texts without length restrictions.