AI Provider Setup

Configure OpenAI, Anthropic, LM Studio, or Ollama for AI-powered metadata generation

← Back to Support

AI Provider Options

Loki Studio uses AI to generate video titles, descriptions, and tags. You have several options depending on your needs, budget, and preferences.

Cloud Providers

  • OpenAI- Fast, affordable
  • Anthropic Claude- High quality

Local Providers

  • LM Studio- Free, easy
  • Ollama- Free, advanced

Option 1: OpenAI (Cloud - Easiest)

Pros & Cons

Pros

  • Easy setup
  • Fast and reliable
  • High-quality output
  • No local resources needed

Cons

  • Requires API key
  • Costs money (~$0.01-0.05 per video)
  • Requires internet connection

Setup Instructions

  1. Get OpenAI API Key
    • Visit:Https://platform.openai.com/api-keys
    • Sign up or log in
    • Click"Create new secret key"
    • Name it:Loki Studio
    • Copy the key (starts withSk-...)
    • IMPORTANT:Save this key securely - you won't see it again!
  2. Add Credits to Account
  3. Configure Loki Studio
    • Open Loki Studio
    • Load your configuration file
    • Navigate toMetadataTab
    • Click"AI Settings"
    • Set:
      • Provider: Openai
      • Model: Gpt-4o-mini(recommended)
      • API Key:Paste your key
    • Click"Save"
  4. Test Connection
    • Click"Test API Connection"
    • Should show: ✅"OpenAI API connected successfully"

Recommended Models

  • Gpt-4o-mini- Fast, cheap ($0.15/$0.60 per million tokens) -RECOMMENDED
  • Gpt-4o- Higher quality, more expensive ($2.50/$10 per million tokens)
  • Gpt-4-turbo- Legacy model, good balance

Cost Estimate

  • 1 video metadata generation: ~$0.01-0.05
  • 100 videos per month: ~$1-5/month

Option 2: Anthropic Claude (Cloud - Alternative)

Pros & Cons

Pros

  • High-quality output (often better than GPT)
  • Longer context windows
  • Good for complex descriptions

Cons

  • Requires API key
  • Costs money (similar to OpenAI)
  • Requires internet connection

Setup Instructions

  1. Get Anthropic API Key
  2. Add Credits
    • Add payment method
    • Purchase credits ($5-10 minimum)
  3. Configure Loki Studio
    • Navigate toMetadataTab >AI Settings
    • Set:
      • Provider: Anthropic
      • Model: Claude-3-5-sonnet-20241022
      • API Key:Paste your key
    • Click"Save"

Recommended Models

  • Claude-3-5-sonnet-20241022- Best balance
  • Claude-3-5-haiku-20241022- Faster, cheaper

Option 3: LM Studio (Local - Free)

Pros & Cons

Pros

  • Completely free
  • Works offline
  • Privacy (data never leaves your computer)
  • No API limits

Cons

  • Requires powerful computer (16+ GB RAM recommended)
  • Slower than cloud providers
  • Quality depends on model chosen
  • Setup more complex

Setup Instructions

  1. Download LM Studio
  2. Launch LM Studio
    • Open LM Studio
    • You'll see the model library
  3. Download a Model

    Recommended Models (sorted by quality):

    Mistral 7B Instruct(Good balance)

    • Size: ~4 GB
    • RAM: 8 GB minimum

    Llama 3.1 8B Instruct(Better quality)

    • Size: ~5 GB
    • RAM: 10 GB minimum

    Llama 3.1 70B Instruct(Best quality)

    • Size: ~40 GB
    • RAM: 48 GB minimum (quantized versions smaller)

    Qwen 2.5 14B Instruct(Good for creative content)

    • Size: ~8 GB
    • RAM: 16 GB minimum
    1. Click on the model
    2. Choose quantization level:
      • Q4_K_M- Good balance (recommended)
      • Q5_K_M- Better quality, larger
      • Q8_0- Best quality, largest
    3. Click"Download"
    4. Wait for download (5-30 minutes depending on size)
  4. Load the Model
    • Click"Local Server"Tab in LM Studio
    • Select your downloaded model
    • Click"Start Server"
    • Server should start onHttp://localhost:1234
    • Keep LM Studio running while using Loki Studio!
  5. Configure Loki Studio
    • Open Loki Studio
    • Navigate toMetadataTab >AI Settings
    • Set:
      • Provider: Lm-studio
      • LM Studio URL: Http://localhost:1234/v1
      • Model:(leave as default or set to your model name)
    • Click"Save"

Test Connection

  • Click"Test API Connection"
  • Should show: ✅"LM Studio connected successfully"

Tips for LM Studio

  • Keep LM Studio open while using Loki Studio
  • If model is slow, try a smaller model (7B instead of 13B/70B)
  • GPU acceleration in LM Studio helps (enable in LM Studio settings)
  • First generation is slow (loading model), subsequent ones faster

Option 4: Ollama (Local - Free, RECOMMENDED)

🎉 NEW: One-Click Model Install!

Loki Studio v1.0.0 now supports installing Ollama models directly from the UI. No command line needed!

Pros & Cons

Pros

  • Completely free
  • Works offline
  • Privacy - data never leaves your PC
  • One-click model install from Loki Studio!
  • Runs as background service (always ready)
  • Lighter than LM Studio

Cons

  • Requires decent GPU (8GB+ VRAM recommended)
  • Slower than cloud providers (5-10x)
  • Initial download is large (3-8GB per model)

Setup Instructions

  1. Install Ollama

    Windows:

    • Visit: ollama.com
    • Download Ollama for Windows
    • Install - Ollama starts automatically as a background service
  2. Option A: Install Models from Loki Studio (EASIEST!)
    1. Open Loki Studio
    2. Go to Settings → Application Settings → LLM Providers
    3. Select Ollama as provider
    4. Click Install next to any model (llama3, mistral, qwen, etc.)
    5. Wait for download to complete (progress shown)
    6. Click Refresh to see installed models
  3. Option B: Install via Command Line (Advanced) 
    # Best balance of quality and speed (8GB VRAM)
ollama pull llama3:8b

# Faster, smaller footprint (4GB VRAM)
ollama pull llama3.2:3b

# Good alternative (8GB VRAM)
ollama pull mistral:7b
  4. Configure Loki Studio
    • Go to Settings → Application Settings → LLM Providers
    • Set AI Provider to Ollama
    • Select your model from the dropdown (auto-detected)
    • Settings auto-save!

Tips for Ollama

  • KeepOllama serveRunning in background
  • List available models:Ollama list
  • Remove models:Ollama rm < model>
  • Check GPU usage:Ollama ps
☕ Buy me a coffee