Migrate from ElevenLabs to Lero

Introduction

If you use ElevenLabs and your audience is in Brazil, this guide shows what changes when you switch to Lero: the API call, the voices, the limits, the MCP and the price. The switch takes a few minutes: one route, one header and the voice name.

Jump ahead: Concepts · Generate speech with the API · Voices · Limits · MCP · Voice agents · Pricing · Checklist

When to switch

• Your audience is Brazilian and the accent matters: Lero has carioca, paulista and niteroiense voices from native speakers.
• Your scripts have numbers, BRL prices, phone numbers, dates and acronyms: Lero reads them the way Brazilians say them.
• You want to pay less: Lero monthly plans cost about half.
• If you need many languages in one project, dubbing or music, ElevenLabs is still the option: you can use both.

Concepts

Generate speech with the API

Before, with ElevenLabs:

curl -X POST "https://api.elevenlabs.io/v1/text-to-speech/$VOICE_ID" \
  -H "xi-api-key: $ELEVENLABS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"text": "Seu pedido chega amanhã.", "model_id": "eleven_multilingual_v2"}' \
  --output speech.mp3

After, with Lero:

curl -X POST "https://app.lero.studio/api/v1/tts" \
  -H "Authorization: Bearer $LERO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"texto": "Seu pedido chega amanhã.", "voz": "rio"}' \
  --output speech.mp3

The response is the MP3 (audio/mpeg), as with ElevenLabs. With Accept: application/json, the audio comes base64-encoded in the audio field. The X-Creditos-Custo, X-Creditos-Restantes and X-Duracao headers carry the cost, the balance and the duration.

Voices

Replace each voice_id with a Lero voice:

• rio: Rio de Janeiro accent.
• sp: São Paulo accent.
• niteroi: Niterói accent.

To choose, generate the same sentence in all three and listen with your real script. GET /v1/voices returns the current list.

Limits

• Up to 500 characters per request. For long text, split on sentences, generate each part and join the MP3s in order.
• The /fala route returns the whole MP3, with no streaming.
• On 429 (muitos_pedidos or ocupado), respect Retry-After. On 503 carregando, retry in a few seconds.
• The API is on the Creator and Dev plans.

MCP

If you generate audio from Claude Code, Codex or Cursor with the ElevenLabs MCP, switch to the Lero MCP:

claude mcp add --env LERO_API_KEY=$LERO_API_KEY lero -- npx -y lero-mcp

codex mcp add lero --env LERO_API_KEY=$LERO_API_KEY -- npx -y lero-mcp

The tools are lero_fala, lero_vozes and lero_saldo. MP3 files are saved to ~/Downloads/Lero.

Voice agents

ElevenLabs conversational agents map to the Lero Conversational Agent, in beta: GET /v1/agents for discovery and POST /v1/agents/conversation to open the WebRTC conversation, with your system prompt and your knowledge. Walkthrough at lero.studio/docs/en.

Pricing

ElevenLabs prices checked on 2026-11-10 at elevenlabs.io/pricing, reference rate US$ 1 = R$ 5,42 + IOF de 3,5%. Plans change: check before deciding. Lero pricing is at lero.studio/en/pricing.md.

Checklist

1. Sign up at app.lero.studio and create a key under Developers.
2. Store the key as LERO_API_KEY.
3. Change the URL, the header and the fields (text → texto, voice_id → voz).
4. Split texts over 500 characters.
5. Generate the same sentence in the three voices and choose.
6. Switch the MCP, if you use it.