FlopCoin Generation Platform

FlopCoin turns a prompt into finished media. You submit a generation job (text, image, video, or audio), a fulfillment worker routes it to the right model or provider, and you poll the job until a result is ready.

Build status is honest here. Only the text lane is live in production today. The gpt-image image lane is code-complete and waiting on provider credentials. Every other lane on this site is PLANNED and documents the intended contract and tuned settings, not a shipping endpoint. Each model is tagged so you always know what runs.

How a job flows

  1. Submit a request with a mode, a prompt, an optional model, and optional image attachments. The job is captured into the generation_requests queue with status queued.
  2. A worker claims the job (status running) and either runs it inline (text via ollama, gpt-image via Azure) or submits it to an async GPU provider and tracks the provider job id.
  3. You poll the job until status is done (read result) or failed (read error_code).

Start here

Generation API

Submit a job, poll status, and read the result shape - accurate to the live pipeline.

Model catalog

Every lane and model, its tunable settings, and whether it is LIVE or PLANNED.

Error codes

Every failure code, what it means, and how to resolve it.

Providers

Runway, gpt-image (Azure), and the open self-hosted models, with links to the official docs.

Modes

ModeStatusWhat it does
textLIVEPrompt to text via an ollama-compatible endpoint (default llama3.2:3b). Fulfilled inline by the worker.
imageCODE-COMPLETE / PLANNEDPrompt to image. gpt-image-* via Azure is code-complete (pending creds); the open RunPod models are planned.
videoPLANNEDPrompt / image to video via an async GPU provider. Provider not yet configured in production.
audioPLANNEDTTS, voice-clone, music, and SFX. The audio mode is not yet in the pipeline's mode enum - see the catalog.

The pipeline's current mode enum is text | image | video. The audio mode and a public keyed HTTP endpoint are on the roadmap; see the Generation API page for exactly what is implemented today.