Vocal Maker

1. 01

   Write the spoken script

   Type or paste up to 5,000 characters per generation (10,000 when signed in). Punctuation steers phrasing, use commas for breath beats and em dashes for harder cuts.

2. 02

   Pick a vocal voice

   54 Kokoro voices across 9 languages. Choose by character: dramatic narrator, smooth female lead, punchy delivery, or calm reader.

3. 03

   Generate the take

   Output is a clean 24 kHz mono WAV by default, or MP3 if you pick it in the format menu. The model is deterministic, same script and voice produce the same take, every time.

4. 04

   Drop the WAV into your DAW

   Import into Ableton, Logic, FL Studio, Reaper, or Audacity. Treat as a dry vocal stem: EQ, compress, add reverb or delay, then align to your tempo grid.

Want to hear them? Browse all 54 voices →

• 01

  Generate clean, handle silence in the DAW

  Kokoro outputs spoken vocals without simulated breath noise, but it does insert micro-pauses on punctuation. Generate the take, then in Ableton or Logic strip-silence the head and tail to a clean -60 dB floor. Use a noise gate at -50 dB threshold with a fast 5 ms attack to remove any tail artifacts before you process the vocal.

• 02

  Treat the WAV as a dry stem

  The 24 kHz WAV is dry, no reverb, no compression, no EQ baked in. That is exactly what you want for production. Run it through your standard vocal chain: high-pass at 80 Hz, presence boost around 3 kHz, de-ess at 6–8 kHz, then 3:1 compression with -18 dB threshold. Add reverb and delay last as sends, never on the dry channel.

• 03

  Layer with your music, sidechain the bed

  When the spoken vocal sits over a beat, sidechain-compress the music bus to the vocal track at 4:1 with a 50 ms release. Music ducks 3–4 dB on every word and rebounds in gaps, instant clarity without manual volume automation. Standard trick for podcast intros, hip-hop skits, and lo-fi spoken samples.

• 04

  Align to a tempo grid with warp / time-stretch

  TTS does not snap to a BPM. To lock the vocal to a 90 BPM beat, drop the WAV into Ableton (Warp → Complex Pro), Logic (Flex Time → Slicing), or FL Studio (Newtone). Mark transients on stressed syllables and snap them to bar-line eighth notes. For speed changes, use Complex Pro at ±15% before formant artifacts appear.

• 05

  Fix phrasing with respelling and punctuation

  There is no SSML support, but punctuation is your steering wheel. Commas are a 100 ms pause, periods are 250 ms, em dashes are sharper cuts, ellipses produce a clear beat. If a word lands wrong, say "data" comes out British when you want US, respell it phonetically ("day-tuh") and regenerate. The model follows the spelling, not a dictionary.

• 06

  Splice long takes with 100 ms crossfades

  For a 20-minute audiobook chapter or a long monologue, generate in 4,000-character chunks at the same voice and speed. Butt-join the WAVs on a single track in your DAW, apply a 100 ms equal-power crossfade at each splice. Because Kokoro is deterministic per voice, splice points are clean, listeners hear one continuous take.