StreamPod Quickstart Guide

Getting Started

Read the guidelines and follow the rules

  1. Read the #rules channel
  2. Follow the guidelines and rules for using StreamPod AI

Creating a Private Chat

  1. Navigate to the #⁠model-chat channel
  2. Type /model to select your preferred AI model
  3. Enter the newly created private thread
    • Thread names are random.
    • Only the user can see the thread creation message(ephemeral threads)
    • Additional users must be invited using the @ symbol

Basic Commands

Text Generation

  • Use /model to access text-based models
  • Optionally set a system prompt to define the model’s role and behavior
  • Model might refuse to load if the system prompt exceeds the character limit or if the system prompt does not meet the guidelines

Media Generation

  • /text-to-image: Generate images from text descriptions
  • /text-to-video: Create videos from text prompts
  • /image-to-video: Convert images to videos
  • /text-to-sfx: Generate sound effects and music

Voice Agent Features

  • /createvoice: Initialize voice chat
  • /talk: Begin voice interaction
  • /leave: Exit voice chat

Usage Guidelines

Text and Vision Models

  1. Select your model with /model
  2. Set an optional system prompt plus optional parameterized settings and enter the private thread
  3. Invite additional users using the @ symbol
  4. Enter your prompt or upload media
  5. Wait for the model’s response
  6. Follow up with additional prompts or commands as needed

Media Generation

Image Generation

  • Maximum prompt length: 1024 characters
  • Maximum image count: 3
  • Pre-prompting supported by models: Claude, Llama, and specialized Vision models

Video Generation

  • Maximum prompt length: 512 characters
  • Image-to-video requirements:
    • Resolution: 1280x720
    • Format: JPEG or PNG (non-transparent)
    • Compression recommended

Best Practices

  • For media generation, pre-prompt multimodal models with context and specific details about the desired output. Details such as the character length, image count, shot type and resolution are crucial for optimal results.

  • Use system prompts to guide the model’s role and behavior

  • Invite additional users to private threads using the @ symbol within the private thread

  • Image generation is parameterized by the model’s input requirements and guidelines, including character limits.

  • Use the image generator’s parameters such as resolution, image count, prompt strength, and other model-specific parameters to optimize the output quality

  • Keep track of your active threads in #⁠model-chat

  • Reference help commands for detailed guidance:

    • /model_help: Text model assistance
    • /media_help: Media generation help
    • /audio_help: Audio feature support

Getting Help

If you need assistance:

Advanced Tips

  • Use the channel menu to track private threads (indicated by a blinking dot)
  • Access thread names through the context menu (right-click)
  • For video understanding tasks, specifically use Nova Lite or Nova Pro models
  • Remember that all video and image generation requests are processed in a queue system and include workflow analytics
  • Use /leave to properly end voice chat sessions and avoid unnecessary resource consumption
  • Models might refuse to process requests if the prompt exceeds the character limit or if the model is busy processing other requests
  • Models might refuse to process requests if the prompt does not follow the model’s input requirements and guidelines
  • To avoid rate limits we recommend using lower parameter count models(Llama 1b & 3b) for general text generation tasks and higher parameter count models for specialized tasks such as multilingual translation and summarization tasks
  • Lower parameter count models might hallucinate often or generate irrelevant content if the prompt is not specific enough or if the prompt is too vague