AI Voice Generator - Text to Natural Voiceover
AI voice generator on Eleven v4: 54 character voices plus 5,000+ community voices, 90+ languages, SRT subtitles. Try free with no sign-up; from $3.90 one-time.
最終更新: 2026-10-18
VoiceSmiths is an AI voice generator that turns a script into a publishable voiceover in about a minute, using ElevenLabs' newest model, Eleven v4. Choose from 54 curated voices, 24 of them original characters, or 5,000+ community voices, and every file comes with subtitles timed to the character. Try it below with no account; paid use starts at $3.90 one-time.
Eleven v4を今すぐ試す：無料・登録不要
1日2回まで無料、1回最大200文字。[角括弧]内のタグは演技に反映され、読み上げられません。
At a glance
| VoiceSmiths AI voice generator | |
|---|---|
| Voices | 54 curated (24 original characters, 30 hand-picked), 5,000+ community voices, cloning, voice design |
| Languages | 90+ on Eleven v4, 70+ on Eleven v3 |
| Direction | Emotion tags like [excited], [whispers], [laughs], performed, not read aloud |
| Output | MP3 + SRT subtitles with character-level timing |
| Free | 2 generations a day, 200 characters each, no account |
| Typical cost | A 30-second English voiceover ≈ 32 credits ≈ $0.27–0.42 on monthly plans |
| Plans | $3.90 one-time for 150 credits (never expire), or from $7.90 a month |
| Payment | Card or Alipay |
| Commercial use | Yes, under the Terms of Service |
Prices as of October 2026.
What is an AI voice generator
An AI voice generator synthesizes near-human speech straight from text. No studio, no scheduling voice actors — paste a script, pick a voice, click generate, and you have publishable audio within a minute.
VoiceSmiths goes one step further: every generation ships with character-level timestamps. Subtitles aren't estimated — they're accurate to the millisecond each character is spoken, and the exported SRT drops straight into your editor.
What you can make with an AI voice generator
- Short-video voiceovers: scripts for TikTok, YouTube Shorts and Reels, localized into multiple languages from one script
- Audiobook & podcast narration: generate long scripts chapter by chapter, with voice cloning keeping the narrator consistent
- Ad reads & promos: direct delivery with inline emotion tags like
[excited]and[whispers] - Course & explainer audio: change one slide, regenerate one paragraph — never re-record the whole lesson
- Character voices: anime, game NPCs and drama roles from the voice library, or a whole scene with multi-speaker dialogue
- Full-cast audio dramas: paste a story into the audio drama studio and it casts a voice per character
Hear it
Short-video voiceover:
Keynote / ad read:
The same short-video pitch in English:
One AI voice generator, very different registers, in Chinese and English — no tool switching.
Choosing a voice
Before you generate anything, pick the right voice. Each of the 54 curated voices has its own page and sample, so audition a few in the voice library first:
- Shorts and product promos: start with a warm, clear voice like the short-video sample above; for keynotes and ad reads, compare the second sample
- News and explainers: a steady broadcast-style voice keeps attention on the content
- Characters and drama: the 24 original character voices (Monkey King, Ice Queen, British Butler, Trailer Voice and more) were built to perform
- Nothing fits: dig into the 5,000+ community voices, clone your own voice for 30 credits, or design a new one from a text description for 20 credits
Keep one voice per project, so a series or a full audiobook sounds like the same narrator from start to finish.
Emotion tags: direct the read, don't just play it
Eleven v4 and v3 accept direction written straight into the script, such as [excited], [whispers] and [laughs]. The AI voice generator treats each tag as a stage direction: it changes the delivery, is never read aloud, and never shows up in the subtitles. The full list is in the emotion tags cheat sheet.
Habits that make the result sound natural:
- Tag the turning points: add a tag where the mood shifts, listen, then decide whether the line needs more
- Let punctuation pace it: ellipses add pauses and capitals add emphasis
- Fix names before you generate: brand names and jargon that come out wrong go into the pronunciation dictionary once, instead of being patched line by line
- Regenerate only what changed: billing is per character, so redoing one line costs only that line
AI voice generator vs. recording yourself or hiring an agency
| Dimension | VoiceSmiths | Record yourself | Voice agency |
|---|---|---|---|
| Turnaround | 1 minute | Hours | 3–7 days |
| Cost per 30-second clip | About $0.27–0.42 | Gear + time | From $50 |
| Languages | 90+ | Impractical | Priced per language |
| Revisions | Regenerate one line | Re-record | Re-book |
Billing is transparent credits: every generate button shows its price up front, and failed generations refund automatically. Pay with card or Alipay.
Start in three steps
- Sign up with Google or email, then pick the $3.90 pack or a plan (Starter from $7.90 a month)
- Open the Voice Studio, paste your script and pick a voice
- Generate, then download the audio and SRT subtitles
FAQ
Is there a free AI voice generator on VoiceSmiths? Yes, for trying it: two generations a day, up to 200 characters each, with no account, on this page and the homepage. Every voice sample on the site is free to play. Longer or more frequent generation uses credits, from $3.90 one-time.
What does a voiceover cost? A 30-second English voiceover is about 450 characters, which is 32 credits on the standard models: roughly $0.27–0.42 on the monthly plans. Speech is billed at 7 credits per 100 characters (4 on Flash), the price is shown on the button, and failed generations are refunded. See pricing.
Which voices can I use? 54 curated voices with their own pages and samples (24 original character voices such as Monkey King, Ice Queen, British Butler and Trailer Voice, plus 30 hand-picked ElevenLabs voices), 5,000+ community voices, your own cloned voice, or a voice designed from a text description. Browse them in the voice library.
Which languages are supported? 90+ on Eleven v4 and 70+ on Eleven v3, including English, Chinese (Mandarin and Cantonese), Spanish, Japanese, Korean, Portuguese, French and German. Details on text to speech.
Can I use the output commercially? Yes. Audio you generate is yours and may be used in commercial projects under the Terms of Service. The 24 character voices are original designs, not clones of real people.
Which AI model generates the voices? ElevenLabs models through the official API: Eleven v4 (released September 28, 2026), Eleven v3, Multilingual v2 and Flash v2.5. VoiceSmiths is independent and not affiliated with ElevenLabs.
Can it dub videos too? Yes — video dubbing transcribes, translates and re-voices an uploaded video end to end.
Get started
Try a line free at the top of this page to hear Eleven v4. When it sounds right, open the studio and use the AI voice generator for your voiceovers, narration and character reads.