Providers

Synthome connects to multiple AI providers to run models. Each provider requires an API key.

Supported Providers

Provider	Environment Variable	Models
Replicate	`REPLICATE_API_KEY`	Video, Image, Audio
Fal	`FAL_KEY`	Video, Image
Google Cloud	`GOOGLE_CLOUD_CREDENTIALS`	Image, Video
ElevenLabs	`ELEVENLABS_API_KEY`	Audio (TTS)
Hume	`HUME_API_KEY`	Audio (TTS)
OpenAI	`OPENAI_API_KEY`	Transcription correction

Quick Setup

Set environment variables for the providers you need:

# .env or .env.local
REPLICATE_API_KEY=r8_...
FAL_KEY=...
GOOGLE_CLOUD_CREDENTIALS='{"type":"service_account",...}'
ELEVENLABS_API_KEY=...
HUME_API_KEY=...
OPENAI_API_KEY=sk-...

How Providers Work

When you specify a model, you also specify which provider to use:

import { videoModel, imageModel, audioModel } from "@synthome/sdk";

// Replicate for video
videoModel("bytedance/seedance-1-pro", "replicate");

// Fal for lip-sync
videoModel("veed/fabric-1.0", "fal");

// Google Cloud for high-res image generation
imageModel("google/nana-banana-pro", "google-cloud");

// ElevenLabs for TTS
audioModel("elevenlabs/turbo-v2.5", "elevenlabs");

The SDK automatically uses the corresponding API key from your environment.

Per-Request API Keys

Override environment variables for specific requests:

generateVideo({
  model: videoModel("bytedance/seedance-1-pro", "replicate"),
  prompt: "Ocean waves",
  apiKey: "r8_custom_key_here", // Override for this request
});

Provider Selection

Some models are available on multiple providers:

// Same model, different providers
imageModel("google/nano-banana", "replicate");
imageModel("google/nano-banana", "fal");

Choose based on:

Existing accounts: Use providers where you already have API keys
Pricing: Check each provider's pricing page
Rate limits: Different providers have different limits

Overview