Text to Speech: Natural AI Voices, MP3 + SRT Subtitles (Kokoro TTS, No API Key)

Turn text, scripts, articles and Markdown into natural-sounding speech with Kokoro, the open-source AI voice model that tops community TTS rankings for its size. Choose from 41 voices in 7 languages (US and UK English, Spanish, French, Italian, Brazilian Portuguese, Hindi), get MP3, WAV, Opus, M4A or FLAC audio plus SRT/WebVTT subtitles with exact sentence timings, and convert hundreds of texts in one run. No ElevenLabs or OpenAI account, no API key, no setup. You pay only for what you convert: $0.04 per 1,000 characters.

What you can use it for

• YouTube, TikTok and Reels voiceovers: turn a script into a clean voice track plus matching SRT captions for your editor.
• Faceless video channels and shorts automation: generate dozens of voiceovers in one batch from n8n, Make, Zapier or your own code.
• Audiobooks and article narration: read long chapters (up to 200,000 characters per text) and merge them into one file.
• E-learning and product demos: consistent narration for courses, explainer videos and onboarding flows.
• Podcast intros, ads and IVR / phone menus: short prompts in MP3 or WAV.
• Accessibility: listen-to-this-article audio for blogs and newsletters.
• AI agents: give an LLM agent a voice via the Apify API or MCP server.

How to use it

1. Paste one or more texts (each becomes its own audio file), or pass items with an id, voice and speed per text, or link .txt/.md files.
2. Pick a voice, speed and audio format. Optionally blend two voices into your own custom voice.
3. Click Start. Each text becomes one result row with links to the audio file and subtitle files.

Pricing

Billing is per started 100 characters of each text ($0.004), counted after Markdown cleanup, plus a tiny start fee. Subtitles, merged files and texts that fail are free.

Job	Characters	Price
A 60-second YouTube Short script	~900	$0.036
50 product video voiceovers	50 x 600 = 30,000	$1.20
A 10-minute explainer narration	~9,000	$0.36
A full 60,000-word audiobook	~350,000 (split into chapters)	$14.00

That's less than half the price of the other Kokoro Actor on the Store, and a fraction of what ElevenLabs charges per character.

Voices

Language	Voices (female / male)
English (US)	af_heart (best), af_bella, af_nicole, af_aoede, af_kore, af_sarah, af_nova, af_sky, af_alloy, af_jessica, af_river / am_michael, am_fenrir, am_puck, am_echo, am_eric, am_liam, am_onyx, am_adam, am_santa
English (UK)	bf_emma, bf_isabella, bf_alice, bf_lily / bm_george, bm_fable, bm_lewis, bm_daniel
Spanish	ef_dora / em_alex, em_santa
French	ff_siwis
Italian	if_sara / im_nicola
Portuguese (Brazil)	pf_dora / pm_alex, pm_santa
Hindi	hf_alpha, hf_beta / hm_omega, hm_psi

The first letter of the voice sets the language and accent. Voice blending: af_heart:0.6, af_bella:0.4 mixes two (or more) voices of the same language into a new voice that nobody else on your niche is using.

Features

• Natural AI voices from the Kokoro-82M model (Apache-2.0, 8-bit quantized), run inside the Actor on CPU. Your text is never sent to a third-party AI API.
• 5 audio formats: MP3, WAV, Opus (OGG), M4A/AAC and FLAC, with selectable bitrate.
• Exact SRT and WebVTT subtitles: every sentence is generated separately and timed to the millisecond, then split into readable 2-line captions.
• Bulk: up to 1,000 texts and 1,000,000 characters per run; per-text voice and speed overrides.
• Markdown-aware: headings, bold, bullet points, code blocks and link URLs are cleaned up so they are not read aloud (ChatGPT/Claude scripts work as-is).
• Natural pacing: configurable pauses between sentences and paragraphs; speed from 0.5x to 2.0x.
• Merge into one file: optional all-texts.mp3 with merged subtitles, for audiobooks and playlists.
• Sentence timings in JSON for karaoke captions, video editors and word highlighting.
• Budget safe: before each text it checks your run's max cost and skips texts that would exceed it (not charged).

Input example

{
  "texts": ["# Why speed matters\n\nA slow page loses visitors before they ever see your offer. This **tool** checks every page on your site and tells you what to fix first."],
  "items": [
    {"id": "es-promo", "text": "El informe mensual estará listo el próximo lunes por la mañana.", "voice": "ef_dora"},
    {"id": "uk-intro", "text": "Please check the website before the meeting on Thursday.", "voice": "bm_george", "speed": 1.1}
  ],
  "voice": "af_heart",
  "format": "mp3",
  "subtitles": "both",
  "mergeIntoOneFile": true
}

Output example (one item per text)

{
  "id": "text-1",
  "source": "input",
  "voice": "af_heart",
  "language": "English (US)",
  "speed": 1.0,
  "characters": 206,
  "textPreview": "Why speed matters.\n\nA slow page loses visitors before they ever see your offer. This tool checks every page on your site and tells you what to fix first. ...",
  "audioUrl": "https://api.apify.com/v2/key-value-stores/<store-id>/records/text-1.mp3",
  "format": "mp3",
  "fileSizeKB": 207.0,
  "durationSeconds": 13.16,
  "sentenceCount": 4,
  "processingSeconds": 3.6,
  "srtUrl": "https://api.apify.com/v2/key-value-stores/<store-id>/records/text-1.srt",
  "vttUrl": "https://api.apify.com/v2/key-value-stores/<store-id>/records/text-1.vtt",
  "billedUnits": 3
}

SRT file:

1
00:00:00,000 --> 00:00:01,173
Why speed matters.

2
00:00:01,773 --> 00:00:05,421
A slow page loses visitors before they
ever see your offer.

Use it from code, n8n or AI agents

Call it through the Apify API (POST https://api.apify.com/v2/acts/manageable_gully~text-to-speech/run-sync-get-dataset-items), the Apify Python/JavaScript clients, the Apify n8n/Make/Zapier integrations, or as a tool for AI agents through the Apify MCP server. The audioUrl in each result is a direct download link.

FAQ

Do I need an ElevenLabs, OpenAI or Google API key? No. The voice model runs inside the Actor.

Can I use the audio commercially (YouTube monetization, ads, courses)? The Kokoro model and its voices are released under the Apache-2.0 license, which allows commercial use. You are responsible for the text you convert.

Can it clone my voice or a celebrity voice? No, on purpose. You can only use the built-in voices or blend them.

Japanese or Chinese? Not yet. Kokoro needs a special phonemizer for them and the quality isn't good enough to charge for.

How long can a text be? Up to 200,000 characters per text (about 3.5 hours of speech) and 1,000,000 characters per run. Split whole books into chapters.

How fast is it? Roughly 50-75 seconds of processing per 1,000 characters (about 1 minute of speech) at the default 4 GB memory. Long jobs run unattended in the cloud; schedule them or call the API and collect the files when the run finishes.

Can I read a web page aloud? Convert it to clean text first with Web Page to Markdown, then pass the text here.

More tools from WebIntel Tools

• Audio & Video Transcriber (speech to text)
• Web Page to Markdown for LLMs
• PDF & Image OCR to Text/Markdown
• URL to PDF & Screenshot