Skip to content

Content Generation

MIFY provides a dedicated content generation system at /content-gen with three independent execution paths.

PathHow It WorksBest For
MIFY QueueSubmits ComfyUI workflows to external GPU providers (Novita, Vast)High-quality image and video generation
AI ProviderDirect calls to AI providers with sync or SSE streamingText generation, quick image generation
ComfyUI LocalRuns workflows on your local ComfyUI installationCustom models, local development

Navigate to /content-gen to see available generation workflows. Each workflow has a dynamic form based on its parameters — you’ll see different input fields depending on the workflow.

Workflows support 12 input types with 11 UI widgets:

WidgetInput TypeExample
Text fieldStringPrompt text
TextareaLong textDetailed descriptions
SliderNumber (with range)Strength, steps, guidance
Image uploadImage URL or fileReference images
Select dropdownEnum/selectModel selection, style choices
CheckboxBooleanEnable/disable features
Model pickerModel referenceChoose a specific AI model

After submitting:

  • MIFY Queue: Job is queued for GPU processing. You’ll see status updates as it progresses.
  • AI Provider: Response arrives immediately (sync) or streams in real-time (SSE).
  • ComfyUI Local: Runs on your local machine, results collected when complete.

Visit /content-gen/history to see all your past generations. Click any run to see:

  • Generated artifacts (images, videos)
  • Input parameters used
  • Cost and timing information
  • Error details (if failed)

For image/video generation via the MIFY Queue path, jobs are processed by the execution layer:

  • Run types: Image, video, and vision (image analysis)
  • Training: Submit model fine-tuning jobs with datasets uploaded to Cloudflare R2
  • Model registry: Store and reuse trained model artifacts
  1. Navigate to /training/new
  2. Upload a dataset (images/files)
  3. Select a training preset
  4. Submit and monitor progress at /training/[id]

Trained models become available for use in future generation workflows.