- Generate images, video, and audio with thousands of open-source models
- Run popular models like FLUX, Stable Diffusion, and Whisper
- Stream long-running predictions or wait synchronously for fast models
- Train and fine-tune custom models on your own data
Common use cases and example apps
| Example app | Example prompt | Description |
|---|---|---|
| Text-to-image generators | Use Replicate and build an app that turns text prompts into images with FLUX. | Generate images from text prompts on demand. The app sends prompts to a Replicate image model and shows the results to the user. |
| Image-to-video tools | Use Replicate and build a tool that animates a product photo into a short video clip. | Turn still images into short video clips. The app uploads a source image, runs an image-to-video model, and plays back the generated clip. |
| Audio transcription apps | Use Replicate and build an app that transcribes uploaded audio with Whisper. | Transcribe spoken audio into text. The app sends recordings to a transcription model and returns the transcript in-app. |
| Image editing tools | Use Replicate and build a tool that removes backgrounds from uploaded product photos. | Edit and enhance images automatically. The app runs an image-processing model and returns the edited file to the user. |
| Avatar and sticker generators | Use Replicate and build an app that creates custom avatar stickers from a selfie. | Create personalized image content from user uploads. The app feeds a user photo to a generation model and produces styled results. |
| Custom model apps | Use Replicate and build an app that runs our fine-tuned model on customer inputs. | Serve your own fine-tuned models inside your app. The app calls a model you’ve trained on Replicate and shows the output to users. |
How to connect Replicate
Workspace admins and owners can connect Replicate. You can create multiple Replicate connections using different API tokens, which is useful for separating environments (for example, development and production) or managing rate limits. When the connection is created, Replicate becomes available across all projects in the workspace. Anyone building in a project can ask Lovable in chat to link their project to it.Prerequisites
Before connecting Replicate, make sure you have:- A Replicate account
- A Replicate API token
- Lovable workspace owner or admin role
All predictions run through this connector are billed by Replicate based on the model and how long each run takes. Billing and quota are handled directly by Replicate, not Lovable.
Step 1: Get a Replicate API token
A Replicate API token lets your Lovable app authenticate with the Replicate API. To create a Replicate API token:Open your Replicate account
Sign in to Replicate.
Open API tokens
Go to Account settings → API tokens.
Step 2: Connect Replicate to Lovable
You can create multiple connections using different API tokens.Configure the connection
- Display name: name the connection, for example
Replicate Prod. This name is only used inside Lovable to identify the connection. - API token: paste your Replicate API token (it starts with
r8_).
Choose who can access this connection
Under Who can access this connection, decide who in your workspace can use the connection:
- Only you (default): only the person creating the connection can use it and its associated data.
- Invite specific people: only you and explicitly added workspace members can use the connection and its associated data.
- Invite entire workspace: click Invite entire workspace to make the connection available to everyone in your Lovable workspace.
Limitations
The Replicate connector cannot:- Run predictions from client-side code. Calls go through Lovable from a server-side endpoint.
- Refresh or rotate API tokens automatically. To rotate, create a new token in Replicate and update the Lovable connection.
- Support per-end-user Replicate login. Each connection represents a single Replicate account shared across all projects linked to it.
How to unlink projects from a connection
Editors and above can remove specific projects from a connection without deleting the connection entirely. The connection will remain available for other projects. To unlink projects:
When unlinked, those projects will no longer have access to through this connection. If a project needs again, you can link it to any available connection.