🚗 Community Roadmap - Member requests (updated 9/18/2026)

Welcome to Self-Host Hub.

This community is for people building automation infrastructure they own: self-hosted workflows, local AI, and systems you run for yourself or your clients.

WHAT IS STUDIO

Self-hosted workflow automation platform for AI content pipelines. You own the infrastructure, pick your providers, keep your keys.

Four pieces:

• App (API + web UI)

• Workers (general, video, audio, ComfyUI/GPU), as Docker images or pip install studio-workers for native GPU, Apple Silicon included

• Community Catalog (free providers and workflows)

• Plus Catalog (advanced production workflows for members)

Deploy however you want: single Docker container, Docker Compose, Kubernetes, GPU Host, or hybrid GPU setups.

📋 ROADMAP

✅ MVP: faceless video pipeline (OpenAI, Leonardo, JSON2Video)

✅ Self-hosted video rendering (replaces JSON2Video)

✅ Self-hosted image generation via ComfyUI (replaces Leonardo): Flux.1 Schnell and Flux 2 Klein, generation and image editing

✅ Self-hosted TTS voice cloning (Chatterbox, replaces OpenAI TTS): clone a voice from a ~10 second clip

✅ Bring your own ComfyUI workflows: upload a package, test it privately, publish it to your users

🔜 Self-hosted video generation via ComfyUI

Every stage so far eliminated an API bill. The pipeline runs on your own hardware.

📅 UPDATES

• 9/18/26 Studio 1.10.0: workers from outside your install wait for your approval before they can join, Wait for Webhook steps pause the run until the callback arrives, uploads get a configurable size limit, and files that share a name or are replaced keep their own contents.

• 8/30/26 Studio 1.9.0: give each worker machine its own revocable key limited to the queues you choose, trigger workflows from signed incoming webhooks, 14 new providers including Claude, Gemini, Slack, Discord, Notion and Stripe, and narrate in one language while prompting images in another.

• 8/21/26 Studio 1.8.0: the video worker is half the size, runs without a GPU, and transcribes subtitles about 2.3x quicker.

• 8/17/26 Studio 1.7.0: ComfyUI packages find your models wherever ComfyUI Manager installed them, and your provider API keys carry across upgrades.

• 8/11/26 New walkthrough: the complete Studio install on Windows. App plus workers on Windows 11 running WSL, start to finish. Another foundational video every other walkthrough builds on.

• 8/9/26 New walkthrough: the complete Studio install on a Mac. App plus native workers on Apple Silicon, start to finish. The foundation video every other walkthrough builds on.

• 8/4/26 Studio 1.3.0: pick ComfyUI workflows from a dropdown (no more model field), upload your own ComfyUI packages, one worker serves multiple queues.

• 7/31/26 Self-hosted image generation shipped: four ComfyUI packages (Flux.1 Schnell bf16/FP8/GGUF, Flux 2 Klein generate + edit).

• 7/28/26 Native workers on PyPI: pip install studio-workers, GPU workers on Mac Apple Silicon without Docker.

• 7/17/26 Voice cloning TTS shipped (Chatterbox): clone a voice from a ~10 second clip, standard and turbo.

• 7/17/26 Self-hosted video rendering shipped, replacing JSON2Video.

• 3/30/26 Studio released. Faceless video workflow, five providers, release pipeline complete.

✅ WHAT'S SHIPPED

• Faceless video workflow: article in, narrated video out, fully self-hosted

• Self-hosted GPU image generation and editing (ComfyUI, Flux family), including packages you upload yourself with a private test lane

• Voice cloning TTS (Chatterbox): standard and turbo, no data leaves your server

• Provider system: upload a JSON package, get a working integration, no code (any REST API)

• Workflow engine: approval gates, iteration, pause/resume, human-in-the-loop decisions, signed webhook and schedule triggers, Wait for Webhook steps, live status while it runs

• Workflow builder: visual flow editor, input mapping between steps, template variables

• Marketplace: providers, workflows, prompts, and ComfyUI packages, one-click install, tier gating

• Multi-tenant orgs with role-based access, and a clean client-facing view separate from your builder view

• Worker security: each worker machine gets its own revocable key limited to the queues you choose, outside workers wait for your approval, and refused requests are logged

• Safer files: a configurable upload limit, and files that share a name or are replaced keep their own contents

• Run it anywhere: native workers (pip install studio-workers, Mac Apple Silicon or Linux CUDA) or Docker images, local API with cloud GPU

🔜 WHAT'S NEXT

• Generic rerun-with-overrides

• Form builder with shareable public URLs

• Notification delivery channels (email, Slack, Discord)

• Usage tracking dashboard

Any questions? Let me know.

Cheers,

Kawika

2 comments

🚗 Community Roadmap - Member requests (updated 9/18/2026)