What This Actually Is
A voice cloning tool that's cheap enough to build a business on
Fish Audio is an AI voice platform: drop in 10 seconds of someone's voice and it hands you back a clone that can say anything, in over 80 languages, from that same 10-second sample. It already powers voice products at HeyGen, Retell, and Sanas, so this isn't a toy, it's the same engine real companies ship with.
The reason it matters for you specifically: it's priced at roughly a sixth of what ElevenLabs charges for the same job. That gap is what turns "cool AI demo" into "thing I can actually resell or run at volume without the bill eating the profit."
This guide covers both paths. The no-code way (clone a voice and generate speech straight from your browser, nothing to install) and the API way (plug a cloned voice into your own app, agent, or automation). Pick whichever matches what you're trying to do.
Before You Start
What you need
A free Fish Audio account
That's the whole requirement to clone a voice and generate speech in the browser. No install, no credit card.
10 to 90 seconds of clean audio (to clone a voice)
A quiet room, one speaker, no music or background noise. Longer samples (30-60 seconds) clone a bit more accurately, but 10 seconds works.
An API key (only if you're building something)
Skip this entirely if you just want to generate voiceovers by hand. You only need it to wire a cloned voice into your own app, agent, or a batch script.
Step 1
Create your account
Go to fish.audio and sign up (email, Google, or GitHub). You land in the dashboard with the free tier active automatically: 8,000 credits a month, up to 7 minutes of generation, no card required.
That's enough to clone a voice and test everything in this guide before you spend a cent.
Step 2
Clone a voice (no code)
In the dashboard, go to My Voice Library → Custom Voice → Build New Voice. You'll get two options: record directly from your microphone, or upload an existing audio file (MP3 or WAV, under 32MB).
Give it a name, pick public, unlisted, or private, add the audio, and hit create. Processing takes a few seconds. That's a working voice clone.
Keep it private if it's not yours
Only clone a voice you have permission to use. Set the visibility to private unless the person is fine with it being public.
Free tier = personal use only
To use a cloned voice commercially (client work, monetized content), you need a paid plan for full commercial rights.
Step 3
Turn it into speech (no code)
Click Use Voice on your new clone, type or paste whatever you want it to say into the text box, and generate. Write in any of the 80+ supported languages and it speaks in that language, same voice, no separate setup per language.
Download the file and that's it, a finished voiceover in a voice that didn't exist an hour ago. This alone covers most of what a solo creator, agency, or small business needs.
Want To Automate This?
The API path (this is how you build the money plays below)
The dashboard is fine for one-off voiceovers. The moment you want to build a voice agent, generate voicemails in bulk, or wire cloned voices into your own app, you use the API instead. Grab a key from fish.audio/app/developers, install the SDK (pip install fish-audio-sdk), and the whole flow is two calls: clone once, generate as many times as you want.
Copy-paste: clone a voice and generate speech from it
from fish_audio_sdk import Session, TTSRequest
session = Session("YOUR_API_KEY")
# Step 1: turn your audio sample into a reusable voice (do this once)
with open("my_voice_sample.mp3", "rb") as f:
voice = session.voices.create(
title="My Cloned Voice",
voices=[f.read()],
)
print(voice.id) # save this — it's your reference_id from now on
# Step 2: generate speech in that voice, as many times as you want
request = TTSRequest(
text="Hey, this is a quick reminder about your appointment tomorrow at 2pm.",
reference_id=voice.id,
)
with open("output.mp3", "wb") as f:
for chunk in session.tts(request):
f.write(chunk) That voice.id is the whole trick. Save it once, then pass it as reference_id anywhere you need that voice to talk, in your own app, a voice agent platform, or a batch script.
Real People, Real Money
3 ways people are already using this to make money
None of these are hypothetical. Here's who's charging for what right now, and roughly what they're charging.
Voice agents for small businesses
A cloned voice plugged into a phone bot that answers calls, books appointments, and sends confirmations for restaurants, clinics, and salons. Builders are charging $100-1,000 per deployment plus a monthly maintenance fee. This is the same reference_id trick above, wired into whatever voice-agent platform you're already using.
Multilingual voicemails for realtors
Clone the realtor's voice once from a clip they already posted online, then generate a library of price-drop alerts, new-listing announcements, and showing confirmations in every language their buyers speak. Realtors are paying $200-500 a month for this, delivered same day.
Dubbing YouTube and course videos
Clone the creator's own voice, translate the script, and regenerate the audio track in a new language so it still sounds like them, not a stranger. Creators pay $50-500 per video to go global without re-recording a single line.
Steal This
Copy-paste: the realtor voicemail script
Here's the second money play from above, turned into a working script. Clone the realtor's voice once, then this loops through as many languages as their buyer list needs, same voice every time.
Copy-paste: batch multilingual voicemails
from fish_audio_sdk import Session, TTSRequest
session = Session("YOUR_API_KEY")
VOICE_ID = "your-cloned-voice-id" # from Step 2
# One message, delivered in every language your client's buyers speak.
# Fish Audio's model handles 80+ languages from the same voice clone,
# so you're not swapping tools or re-recording anything.
messages = {
"english": "Hi, this is Sarah with Sunview Realty. The house on Maple Street just dropped to $410,000, give me a call if you want to see it.",
"spanish": "Hola, soy Sarah de Sunview Realty. La casa en Maple Street bajo de precio a $410,000, llamame si quieres verla.",
"mandarin": "你好,我是Sunview地产的Sarah。枫树街的房子降价到41万美元了,想看房请给我打电话。",
}
for lang, text in messages.items():
request = TTSRequest(text=text, reference_id=VOICE_ID)
with open(f"voicemail_{lang}.mp3", "wb") as f:
for chunk in session.tts(request):
f.write(chunk) The Honest Part
What it actually costs
The free tier is real and it's enough to test every step above. Here's what the paid tiers look like once you're using this for real client work.
Free — $0/mo
8,000 credits a month, up to 7 minutes of generation, 3 public voice slots. Personal use only, no commercial rights.
Plus — $11/mo
250,000 credits, up to 200 minutes of generation. Enough for a handful of client projects a month.
Pro — $75/mo
2,000,000 credits, up to 1,620 minutes of generation. This is the volume tier for anyone running the money plays above at real scale.
API, pay-as-you-go — $15 per 1M characters
No subscription needed for the API path. $15 covers roughly 180,000 words, about 12 hours of speech. That's the number that makes the money plays above actually profitable.
Time-Sensitive
Get 50% off, but only until August 31
Fish Audio is running an anniversary sale right now: 50% off every Creator plan (Plus, Pro, and Max), from July 24 through August 31, 2026. No promo code, no waitlist, it applies automatically at checkout.
Go to fish.audio/plan, pick a plan, and the 50% off is already baked into the price. That's it.
The deadline is real
This is Fish Audio's own promotion, not something Zach negotiated, so there's no extension coming. If you're going to upgrade to run any of the money plays above at volume, this is the cheapest that plan will be for a while.
Go Deeper
3 videos worth watching
How to Clone Your Voice in 2 Minutes
The fastest possible walkthrough of Step 2 and Step 3 above, if you'd rather watch it than read it.
How to Clone Your Voice Using AI (Tutorial for Beginners)
Goes further into the API and shows a real app built on top of a cloned voice, useful once you're past the basics.
The Laziest AI Side Hustle Anyone Can Start Right Now
The realtor voicemail money play from above, walked through start to finish with real pricing.
The Whole Thing in 5 Steps
Quick recap
Create your account. Free, no card, at fish.audio.
Clone a voice. 10 to 90 seconds of clean audio, recorded or uploaded.
Generate speech. By hand in the dashboard, or through the API to automate it.
Pick your money play. Voice agents, realtor voicemails, or video dubbing.
Grab 50% off. fish.audio/plan, before August 31.
Work with Me
Want me on your problem for an hour?
Book a 1:1 call and we build the fix live — the workflow you want automated, the tool you can’t crack, whatever’s stuck. Direct, hands-on, no pitch waiting at the end.