CastWorkshopGitHubAI Character Builder
CastWorkshop is an open-source, local-first environment for creating persistent AI characters that can move from conversation into voice, images, scenes, and video.
CastWorkshop conversation-to-media experience.
Create a character, talk with them, establish their identity and world, and bring them together with other characters — without rebuilding everything from scratch for each generation tool.
Persistent characters · Local AI · Web + optional Telegram
Product Preview
Follow Sloan and Mateo from character creation into a continuing story, with their identities carried into conversations, scenes, and shared performances.
Watch the CastWorkshop workflow.
Character creation leads into conversation, image, voice, scene, and multi-character video.
Prepare your local AI stack with Python, ComfyUI, and Ollama. Follow the CastWorkshop installation guide for the supported versions and setup steps.
Connected Experience
CastWorkshop treats characters as persistent participants in an evolving world rather than isolated prompts. Their identity, voice, conversations, visual references, scenes, and relationships can carry forward as you create new media.
Create a character with their own appearance, personality, visual reference, and voice.
Their identity stays attached to the character as you continue talking and generating media.
Character profile, identity image, and selected voice.
Place the character into a scene, establish a visual anchor, and continue generating from that environment.
New conversations and media can build around an existing character and setting instead of restarting from zero.
The same character in a recognizable setting, with related outputs.
Introduce another persistent character and generate scenes involving both.
CastWorkshop keeps track of who is present, which visual and voice references belong to each character, and who is speaking during multi-character video.
Separate character references followed by the strongest shared scene.
Character Continuity
Generative media tools are good at creating individual outputs. The harder problem is keeping a character recognizable as their circumstances change.
CastWorkshop keeps character identity references, selected voices, conversations, approved scene anchors, and generated media attached to the same persistent character.
Same character, different circumstances.
A character can move between different settings and interactions without requiring you to manually reconstruct their identity every time.
Technical Foundation
CastWorkshop sits above individual AI models and workflows. It coordinates character state, conversations, model handoffs, long-running jobs, media references, scene continuity, recovery, and delivery into one character-centered experience.
Persistent character state · Conversation continuity · Scene and media relationships · Orchestration · Jobs and recovery · Browser and Telegram experience
Language · Image · Voice · Video
Runs the language model used for character conversation and media planning.
Runs the image, voice, and video workflows used by CastWorkshop.
Generate character references, identity-preserving images, and scene imagery.
Creates and maintains character voice references for speech generation.
Turn character and scene references into single- and multi-character performance.
System Requirements
CastWorkshop coordinates multiple substantial local models and workflows. The suggested starting configuration below is a planning guide, not a validated minimum. Compatibility depends on the underlying AI stack and selected workflows:
Testing has primarily been performed on a Windows 11 system with an NVIDIA GPU. Minimum VRAM, a hardware compatibility matrix, complete fresh-install acceptance, and beta GPU acceptance are not yet verified. Linux and macOS GPU paths remain unverified; offline Python tests do not establish media-generation compatibility.
Image plus video and voice model downloads total approximately 67.6 GB before reuse, excluding Ollama and Whisper models. Allow additional space for ComfyUI, Python environments, caches, temporary files, and generated media. Setup reuses matching filenames and sizes; it does not hash existing models.
Full installation requirements on GitHubRelease Notes
The public source beta is aimed at local-AI users, generative-media creators, developers, and people interested in persistent AI character workflows.
The source includes character, conversation, voice, image, and video workflows, including multi-character media. Install from source; a packaged GitHub Release is not yet published.
This is an early public release. Output quality and compatibility still depend on the underlying models, workflows, GPU environment, and prompts.
CastWorkshop does not claim universal GPU or operating-system compatibility, perfect character consistency, perfect lip sync, or production-ready public deployment.
Create persistent local AI characters and turn them into conversations, voices, images, and video from one connected system.