Experimental creator program · Creator access required to submit
ToonPG is a highly experimental platform for serial webtoons with RPG play. Authors share playable chapters; readers explore, talk and act inside the same story space, on a phone or together on TV.
The story stays linear. Movement, dialogue and short actions happen on one persistent stage. Readers watch automatically and can take over immediately with a joystick or TV remote. Write moments that children, teens and parents can follow together.
You supply YAML, art and sound. The platform supplies movement, dialogue, save/resume and action modules. Begin with one room, one small interaction and a clear ending. Test walking, text size, sound and every resume point before growing the episode.
From a tiny episode to a review receipt
This local QA recording uses a roughly five-second technical episode with existing ToonPG artwork. It demonstrates real rendered playback and an actual private test submission. It is a format demonstration, not a new published story.
The QA run simulates reviewer and administrator roles with temporary test credentials. Live editorial decisions require a human administrator.
Choose the source folder and submit. A receipt shows the immutable revision and compiler, language, bot and resume results.
The reviewer assistant reads the sealed source and inspects images through MCP. Its recommendation and uncertainties appear on the same receipt.
A human editor approves, requests changes or rejects. Approval still needs a source PR and publication checks. The QA test record and temporary playable source were removed after verification.
Package format
Submit one episode folder with the layout below. Use your assigned creator ID and matching series/episode IDs. Both statuses must be draft. English is the source language; complete English, Korean and Japanese display dictionaries are required. IDs and asset paths never change with language.
Limits: 2 MiB per file, 32 MiB total, 256 files, 2,000 steps and 32 checkpoints. Accepted file types: YAML, JSON, PNG, WebP, JPEG, WAV, MP3 and OGG. Use relative local asset paths. Code, HTML, external asset URLs, archives and symlinks are not accepted.
A minimal story beat
# episodes/001/story/main.yaml
beats:
- id: first-step
steps:
- stage: garden
- place: { hero: start }
- control: on
- waitFor: { reach: flower }
- control: off
- dialogue:
- hero: "A little light for the road."
- complete
The referenced stage must define start and the flower zone, and episode.yaml must declare hero in the cast. Include the character sheet, background and cover. The compiler rejects missing references; the bot must reach the ending and resume from every checkpoint.
Locale JSON maps each canonical English display string to its translation. The English value repeats the source string. Include dialogue, names, narration, objectives, stage labels, item text and series/episode metadata. Do not translate YAML keys, IDs or filenames.
{ "A little light for the road.": "A little light for the road." }
Review workflow
Prepare locally and obtain creator access from the operator. The experimental intake is invite-only; there is no self-service account signup.
Read the submission permissions and confirm your file rights. Upload a draft through the web form or MCP, then seal its exact revision.
The server runs the real content compiler, complete language checks, autoplay and checkpoint resume. Failed checks remain visible for feedback and block approval.
An operator-approved reviewer assistant uses MCP to read source and assets, then records a structured advisory report. This is AI-assisted review in the MCP client, not an automatic server-side model call or proof of asset ownership.
A human administrator makes the editorial decision. A review cannot approve itself. Overriding AI concerns requires an explicit acknowledgement; compiler errors cannot be overridden.
For changes requested, make a replacement draft with the earlier submission ID in supersedes. A sealed revision cannot be edited. Keep permanent content IDs stable.
An approved package is exported to a new private source directory for a maintainer PR. Existing metadata and episode IDs must be reconciled in review. Only the normal release process can make it public.
Connect with MCP
Use an HTTP MCP client with the endpoint below and an operator-issued Bearer token. Configuration names and secret substitution differ by client; store the token in its secret settings. Never put tokens in a URL, manuscript or shared screenshot.
Creator, reviewer and administrator tokens expose different tools. Creators can see only their own submissions. Reviewer tools prepare_review, read_review_asset and record_ai_review record advice; decide_submission is administrator-only. Test submissions can never be registered for publication.
Intake stays unavailable until private storage, origin and access keys are configured. An unavailable server returns 503; it does not pretend to accept your manuscript. Save your source folder and retry after the operator opens access.
Your rights
You keep ownership of your work. Submission gives ToonPG the limited permissions stated below for review and approved service delivery. Other authors receive no reuse license. The platform code, original stories and assets remain proprietary, with third-party licenses respected.
Submission permissions · 2026-10-21
You retain ownership of your submission. You confirm that you are authorized to provide every file and to permit review of it. You grant ToonPG a non-exclusive permission to store, copy, build, run and review this submission, including AI-assisted review in the operator-approved MCP client. If an administrator approves it, you permit ToonPG to host, display and deliver the approved episode and its assets as part of the ToonPG service. This does not give other creators a right to reuse your work. Do not submit confidential or personal information that is unnecessary for the story. These permissions do not include unrelated model training. Contact the operator for withdrawal or rights questions; removal of a published episode uses the editorial release process.