Connect an LLM API Key on the Desktop App
Voice Creator Pro uses an LLM for Viral Clips, audiobook casting, and AI Cut. VCP Cloud includes it, and on the desktop app you connect your own provider, pick a model for each feature, and set reasoning and OpenRouter provider order.
Voice Creator Pro uses a large language model (LLM) for Viral Clips, audiobook casting, and AI Cut. VCP Cloud includes an LLM, so there is nothing to set up. On the desktop app you bring your own LLM by connecting a provider in Settings, and you choose the model each feature uses.
Only needed on desktop
This setup is for the desktop app. VCP Cloud has an LLM built in, so if you use Voice Creator Pro in your browser there is no key to add.
Where the LLM Is Used
Each of these features has its own section under AI Models, so you can give each one a different provider and model.
Viral Clips
In Viral Clips, the LLM:
- Finds the moments worth clipping in a long video.
- Makes the auto-edit decisions, such as where to place zoom-ins and which filler words to cut.
- Writes the hook line and the captions and titles to post with each clip.
- Tags each clip as Insight, Storytelling, or Funny.
Audiobooks
In Audiobooks, the LLM reads the manuscript to cast the book:
- Find speakers works out who speaks each line and picks traits for each character's voice.
- Tag emotions marks the mood of each part of the text.
AI Cut
In AI Cut, the LLM finds bad takes, checks your recording against your script, and helps decide which filler words to cut. Bad takes and script checks only run when a model is connected.
Supported Providers
Each section has a provider dropdown with these choices:
- Gemini
- Anthropic
- OpenAI
- Custom endpoint - any OpenAI-compatible API, set by its base URL. Use this for OpenRouter, DeepSeek, or a model running on your own computer.
Connect a Model
1. Open Settings and Go to AI Models
In the desktop app, open Settings and select the AI Models tab. You see one section for each feature: Viral clips, Audiobooks, and AI Cut.
2. Choose a Provider
In the section for the feature you want to set up, pick your provider from the dropdown. Choose Custom endpoint to point at any other OpenAI-compatible API.
3. Paste Your API Key
Enter your provider's key in the API key field. You generate this key from your provider's dashboard, not from Voice Creator Pro.
Keys are saved per provider. If you pick the same provider in another section, it uses the key you already entered.
4. Set the Base URL (Custom Endpoint)
If you chose Custom endpoint, set the Base URL to your provider's OpenAI-compatible endpoint, for example https://openrouter.ai/api/v1.
For a model running on your own computer, such as one served by Ollama at http://localhost:11434/v1, you can leave the API key empty.
5. Choose a Model
Under Model, pick a model from the list your provider offers, for example deepseek/deepseek-v4-flash on OpenRouter. Click the Refresh models button next to it to reload the list. If the list doesn't load, type the model name instead.
Each feature needs its own model, so repeat these steps in each section you plan to use. Changes save on their own, and you see AI settings saved when they do.
Reasoning
Each section has a Reasoning switch. Reasoning uses more tokens and can improve accuracy. It works with Gemini 2.5 Flash and Gemini 3 models, and with OpenRouter on a custom endpoint. For other models the switch is greyed out.
When reasoning is on and the model offers levels, pick a Reasoning effort. The list only shows the levels your model supports, from Minimal up to Maximum. Default leaves it to the model.
Choose OpenRouter Providers
When the Base URL points at OpenRouter, a Providers field appears. OpenRouter can serve the same model from several providers, and they differ in speed and reliability. To choose which ones serve your requests:
- Type a provider name, such as
baidu/fp8. - Press Enter or click Add provider.
- Add more in the order you prefer them. Each one is numbered by its place in the list.
Requests only go to the providers in your list, tried in that order. Click the × on a provider to remove it.
Set Up from a Feature
You can also get to these settings while you work:
- Viral Clips - The Review step shows the AI model for the run, and you can switch models there. If nothing is connected, click Edit on that row to open Settings.
- Audiobooks - If no model is connected when you start Find speakers or Tag emotions, the dialog shows the same provider, key, and model fields, with Open settings for the full tab.
- AI Cut - If bad takes or script checks are turned on without a model, setup shows Open settings.
Notes
- Billing is with your provider. Your API usage is billed directly by the provider whose key you connect, based on their pricing. Voice Creator Pro does not add a charge on top.
- Keep your key private. Your API key is stored for the app to use. Treat it like a password and rotate it from your provider's dashboard if it is ever exposed.
- Prefer VCP Cloud instead? VCP Cloud runs the same features with a built-in LLM, so you can skip provider setup entirely.
Next Steps
- Viral Video Clips - Turn a long video into captioned clips
- Voice Assignment - Give each character in your audiobook a voice
- Video Dubbing - Dub your clips into another language
Viral Video Clips
Drop in a long video and let AI find the moments worth clipping, then cut them into captioned clips for TikTok, Reels, and Shorts in vertical, square, or landscape, with auto-editing or full manual control.
AI Cut
Upload a talking-head video, podcast or lesson and AI Cut removes long pauses, filler words and bad takes. Review every cut, add styled subtitles, change words with Overdub, and export a video, an XML for Premiere or Resolve, or an SRT file.