Free, natural AI voiceovers for your videos (no subscription)

The robotic built-in voices kill a reel. Here's how to get a natural, human-sounding voice for free — the exact setup StackScout uses.

TL;DR: use edge-tts (free, neural, no signup) for instant results, or Kokoro-TTS (free, local, open-weights) when you want the most natural voice and full offline control.

---

Option A — edge-tts (fastest, free, no account)

edge-tts taps Microsoft Edge's neural voices. No API key, no signup. Outputs mp3.

Install (macOS/Linux/Windows, needs Python 3.8+):

pip install edge-tts        # or: pip3 install --user edge-tts

Generate a voiceover:

python3 -m edge_tts \
  --voice en-US-AndrewNeural \
  --rate="+8%" \
  --text "This is the exact voice I use for my videos. Sounds human, costs nothing." \
  --write-media vo.mp3

Best voices (conversational, natural):

List everything: python3 -m edge_tts --list-voices

Tips

• --rate="+8%" speeds it up a touch (reels want energy). --pitch="+2Hz" to brighten.
• Write short sentences. Punctuation = pauses. Spell tricky words phonetically ("stackscout dot app", "M C P").
• One clip per scene → easier to time. Measure length with afinfo file.mp3 (mac) or ffprobe.

---

Option B — Kokoro-TTS (most natural, free, fully local)

Open-weights TTS model, runs on your machine (even CPU). No limits, no internet after setup. Best quality of the free options.

pip install kokoro soundfile

from kokoro import KPipeline
import soundfile as sf
pipe = KPipeline(lang_code="a")            # 'a' = American English
audio = next(pipe("Your script here.", voice="af_heart"))[2]
sf.write("vo.wav", audio, 24000)

Voices: af_heart, af_bella, am_michael, am_puck, … Swap voice= to taste.

---

Where this fits in a video pipeline

1. Write the script (one line per scene).
2. Generate one voice clip per scene (edge-tts loop).
3. Drop the clips into your editor / Remotion timeline, time visuals to the audio.
4. Add subtle SFX (whoosh on cuts, a soft "ding" on reveals) — search free packs.

Paid upgrade (optional)

If you go viral and want the absolute best: ElevenLabs (~€5/mo) has the most expressive voices and voice-cloning. Swap it in later — same 1-clip-per-scene workflow. Not needed to start.

---

Guide by StackScout · more free AI-dev guides, skills & repo picks → stackscout.app