> For the complete documentation index, see [llms.txt](https://themochi-app.gitbook.io/documentation/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://themochi-app.gitbook.io/documentation/module-4-ai-setup-and-intelligence/voice-cloning.md).

# Voice cloning

## Voice Cloning: Send AI Voice Notes in Your Own Voice

**What it is** — Voice Cloning connects an ElevenLabs voice clone of the creator's actual voice to Mochi, so any setter can send a personalized-sounding voice note without the creator recording it themselves. Voice notes convert at roughly 3–4x the rate of plain text in high-ticket DM sales, and a well-made clone is difficult for a lead to distinguish from the real thing.

**Who can use it** — Connecting and configuring the clone lives at `Settings > Intelligence > Voice Clone`, available on **both web and iOS**. Expected to be a **Creator/Manager** action to connect; once connected, any setter with voice-note access can use it when messaging leads (see Module 03).

**Before you start**

* An ElevenLabs account on their **Professional plan** (roughly $11/month — billed directly by ElevenLabs, a separate cost from Mochi) or higher, the tier needed to create a Professional Voice Clone.
* 90–120 minutes of clean recorded audio (60–90 minutes minimum).
* A quiet room and, ideally, a dedicated external microphone rather than your phone or laptop mic.

### Why voice notes convert

Text is easy to skim past. A voice note forces attention and carries tone and warmth that text can't, while still feeling personal even when it's sent at scale. In high-ticket DM sales specifically, that difference shows up directly in reply and conversion rates — which is why cloning the creator's actual voice, rather than relying on a generic AI voice, matters.

### Step 1: Record your source audio

* Record in a **quiet room** — no background noise, no echo.
* Speak in a **natural, conversational tone** — imagine you're actually talking to one lead, one-on-one.
* Use **real variety in pace and energy**. Flat, monotone narration produces a flat, monotone clone.
* **Do not use course or teaching recordings** as your source. That "teaching voice" clones the way it sounds — polished and a little performative — which comes across as stiff and unnatural in a DM.
* Trim out obvious stutters and long pauses before uploading.
* Use a dedicated external microphone if you have access to one; even an inexpensive one beats a phone or laptop mic.
* Aim for 90–120 minutes total. Treat 60–90 minutes as the floor, not the target.

### Step 2: Create a Professional Voice Clone in ElevenLabs

1. Go to **Voices > Create a Voice > Professional Voice Clone** in ElevenLabs.
2. Upload your recorded audio.
3. Wait for the clone to finish processing before moving on — this can take a few hours on ElevenLabs' side. ElevenLabs shows a **yellow tick** once processing is complete; don't try to connect the voice to Mochi before it appears.

### Step 3: Get your API key and Voice ID

1. Go to **Developers > API Keys** in ElevenLabs and click **Create Key**.
2. Enable both the **Text to Speech** and **Models** scopes on the key, then copy it. Both scopes are required, or the connection will fail.
3. Go to **Voices > My Voices**, open your new clone, and copy its **Voice ID**.

### Step 4: Connect in Mochi

**Web and iOS:**

1. Go to `Settings > Intelligence > Voice Clone`.
2. Paste your ElevenLabs **API key** and **Voice ID**.
3. Click **Save**.
4. Confirm it shows **Voice Connected**.

### Step 5: Test and tune

Type a test message, convert it, and listen back. Adjust the **stability** and **speed** sliders directly inside Mochi to get the tone right. ElevenLabs' own slider settings inside their platform do not carry over automatically — re-match them in Mochi itself.

### Sending a cloned voice note

Once connected, a setter can type a message, click the voice-note icon, preview the generated audio, and send it as a native Instagram voice note. Layering in a little ambient background sound before sending can make it feel even more natural and less obviously generated. See Module 03 for the full messaging mechanics.

### Troubleshooting

| Issue                                 | Fix                                                                                                                  |
| ------------------------------------- | -------------------------------------------------------------------------------------------------------------------- |
| Voice sounds robotic                  | Re-record with more natural variety in pace and energy; lower the stability slider in Mochi.                         |
| Sounds too polished, hyped, or formal | Your source audio was likely a course or teaching recording — re-record in a conversational tone.                    |
| "Voice not connected" error           | Check that your ElevenLabs API key has both Text to Speech and Models scopes enabled; re-paste the key and Voice ID. |
| Poor clone quality overall            | Usually background noise in the source recording — re-record in a quieter environment.                               |
| "Voice limit of 1 reached"            | This is an ElevenLabs account-tier restriction, not a Mochi error — check your ElevenLabs plan.                      |

### Tips

* Record in more than one sitting if needed — total minutes matter more than doing it in one take.
* Re-listen to a sample of your source audio before uploading; if it sounds flat to you, it will clone flat.
* Revisit your stability/speed settings after the first week of real sends — what sounds right in a test often needs a small tweak once you hear it in real conversations.

### Common mistakes & fixes

| Mistake                                                  | Fix                                                                                  |
| -------------------------------------------------------- | ------------------------------------------------------------------------------------ |
| Using an old course or webinar recording as source audio | Re-record fresh, conversational audio — teaching audio clones a stiff, formal voice. |
| Enabling only one of the two required API scopes         | Both Text to Speech and Models must be enabled on the same key.                      |
| Assuming ElevenLabs' slider settings apply in Mochi      | They don't carry over — set stability and speed inside Mochi directly.               |
| Recording too little audio                               | 60–90 minutes is the floor; 90–120 gives noticeably better results.                  |

### Limits & good to know

* The ElevenLabs plan required for a Professional Voice Clone is billed directly by ElevenLabs, separate from Mochi.
* Voice clone processing happens on ElevenLabs' side and can take a few hours after upload.
* Setup nav is `Settings > Intelligence > Voice Clone` — some internal SOPs describe it as Organization/Organisation instead; see the NEEDS VERIFICATION note above.

**Related:** Set Up Your AI: Setter Playbook & Creator DNA, Lead Summary. Also see **Voice Notes & Content Hub** and **Send Images, Videos & Files** (Module 03).