07.11.2026

Ukrainian text to speech by API - natural voices, MP3 and subtitles

Turn Ukrainian text, articles or DOCX/PDF files into natural speech with one API call - three Ukrainian neural voices, MP3 plus SRT, a real run you can listen to, the price and the limits.

Good Ukrainian voices are rare in text-to-speech APIs: many offer none, and open voices often put the stress on the wrong syllable. This guide shows one API call that turns Ukrainian text into an MP3 with subtitles, with a sample you can play right here. The run below is from 7 November 2026.

Disclosure: I built the tool used here, Text to Speech on Apify. The voices run inside it (Piper neural models), so no other service and no key of yours is involved. The source code is open: AndriyVK/-apify-document-tools.

Listen first

Voice uk_UA-tetiana-high, speed 1.12, MP3. This is the file exactly as the API returned it, not edited.

The call

from apify_client import ApifyClient

client = ApifyClient("<YOUR_APIFY_TOKEN>")
run = client.actor("ai_admin/text-to-speech").call(run_input={
    "text": "Добрий день! Це український голос, створений нейромережею.\n\n"
            "Він читає статті, документи та субтитри природно і без акценту.",
    "voice": "uk_UA-tetiana-high",   # or uk_UA-mykyta-high, uk_UA-oleksa-high
    "speed": 1.12,
    "format": "mp3",                 # mp3, wav or ogg
    "subtitles": True,               # SRT and VTT, free
})
item = next(client.dataset(run["defaultDatasetId"]).iterate_items())
print(item["durationSeconds"], item["audioUrl"], item["srtUrl"])

Instead of text you can pass sources: links to TXT, Markdown, HTML, DOCX or PDF files (with a text layer), one audio file each.

The real output

voice: uk_UA-tetiana-high   characters: 120   sentences: 3   durationSeconds: 10.59

And the subtitles that came with it - one cue per sentence, timed to the audio:

1
00:00:00,000 --> 00:00:01,068
Добрий день!

2
00:00:01,291 --> 00:00:04,240
Це український голос, створений нейромережею.

3
00:00:04,865 --> 00:00:10,589
Він читає статті, документи та субтитри природно і без акценту.

The whole run took 30 seconds, most of it the container start. As a rule of thumb, 750-900 characters make one minute of speech.

Price

Per started minute of audio: $40 per 1,000 minutes with the high voices (4 cents a minute) and $10 per 1,000 with medium/low voices. All three Ukrainian voices are high. The sample above is one started minute - 4 cents. Subtitles are free.

Limits

• Numbers, abbreviations and foreign words are read by the voice's rules and may sound odd: write «двадцять п'ять» or check how «25» sounds.
• Scanned PDFs without a text layer are not read - run them through OCR first (see Ukrainian scans to Markdown).
• It is a synthetic voice. For an audiobook it is fine for drafts and accessibility; a professional narrator is still better for a final release.

From an AI agent

https://mcp.apify.com?tools=ai_admin/text-to-speech adds the voice as an MCP tool, so an agent can write a text and hand back the audio.