# Infinicia

> Infinicia is Infinite People's AI creation studio (infinicia.ai): short videos with sound from a sentence and a picture or two, songs written and sung from a sentence or one's own lyrics, pictures edited one region at a time (touch an object, say what changes), made from words, widened or enlarged, presentations written by its own model, voices kept (recorded or designed from words) that say texts and take over recordings; characters and 3D objects to come. One Infinite People account and one subscription for every Infinite product, projects that keep what is made (pictures, videos, songs, voices and musics) for good, a bench that assembles them into a montage, and a personal API key that does everything the app does.

Infinite People (infinitepeople.ai) is a French company. The studio runs its video model (MiniMax H3), its song model (ACE-Step 1.5), the segmentation (SAM 3), the enlargement (Nomos upscalers), the voices (VoxCPM2) and their conversion (CosyVoice 3) on its own GPUs; the direction help (rewriting a request, painting a first image, writing lyrics), the jury and the picture edits use Alibaba Model Studio and OpenAI models. Every request is moderated before it runs. Creations are kept a year by default (ephemeral 48 hours or permanent, the account chooses); what a project holds never expires; a deletion waits 48 hours in the bin. A creation can be shared with one person by a one-time link, to see or to edit.

## Product

- [Infinicia, the app](https://infinicia.ai/): web and iOS. A new account gets one free week of the Essentiel plan from its first creation; then an Infinite People subscription — Essentiel (5 €/month), Starter (9.99 €), Standard (19.99 €, adds the "Cinema" direction level) or Pro (49.99 €, the priority queue at the normal price) — widens the week and unlocks longer videos and options. Videos of 5, 10 or 15 seconds, 16:9, with dialogue in any language.
- Trades: Video (available), Song (available), Image (available: edit, generate, extend, enlarge), Presentation (available), Voice (available: keep a voice, say a text, convert a recording), Character, 3D (coming).

## API

- [API documentation (HTML)](https://infinicia.ai/api/): make a video (`POST /v1/video/generate`), poll a job, upload pictures, make a song (`POST /v1/song/write` then `/v1/song/generate`), edit a picture (`POST /v1/image/segment/open` and `/tap` to say where, `POST /v1/image/edit` to say what; `/generate`, `/extend`, `/upscale`), give a voice (`POST /v1/voice/voices` to keep one, `/v1/voice/say`, `/v1/voice/convert`), keep things in a project, assemble a montage on the bench, statuses, refusals, the week (`GET /v1/studio/me`).
- [API documentation (Markdown)](https://infinicia.ai/api/index.md): the same, for readers and language models.
- [OpenAPI document](https://api.infinitepeople.ai/v1/studio/openapi.json): every route a key can reach, machine-readable.
- Base URL `https://api.infinitepeople.ai/v1`, bearer key made in the app (Account > API access). A key reaches the studio's trades only, never the subscription. Feedback: `POST /v1/studio/feedback`.
- [MCP server](https://api.infinitepeople.ai/v1/mcp/crea) — **how to add it and what it does: https://infinicia.ai/mcp/ (and /mcp/index.md)**. Tools named for what a person wants: `get_account`, `set_account`, `list_creations`, `get_creation`, `manage_creation`, `share_creation`, `list_projects`, `get_project`, `manage_project`, `keep_in_project`, `list_project_assets`, `make_video`, `take_frame`, `edit_video`, `ask_song_questions`, `write_song`, `review_song`, `make_song`, `split_song`, `ask_review`, `send_review_version`, `send_review_feedback`, `make_picture`, `edit_picture`, `make_deck`, `rework_deck`, `export_deck`, `present_deck`, `list_voices`, `design_voice`, `say_text`, `convert_voice`, `send_feedback`. The live table with every schema, readable without an account: https://api.infinitepeople.ai/v1/mcp/tools. The server's card: https://api.infinitepeople.ai/.well-known/mcp-server-card (also at https://infinicia.ai/.well-known/mcp). JSON-RPC 2.0 over Streamable HTTP, protocol revisions 2026-07-28 and 2025-11-25. In Claude: Settings > Connectors > Add custom connector with the endpoint (direct link https://claude.ai/new?modal=add-custom-connector#customize/connectors/yours), then sign in with Google (OAuth 2.1: PKCE, dynamic registration, client metadata documents; https://api.infinitepeople.ai/.well-known/oauth-authorization-server). In Claude Code: `claude mcp add --transport http infinicia https://api.infinitepeople.ai/v1/mcp/crea`, or `--header "Authorization: Bearer <key>"` with a key made in the app. In ChatGPT: Settings > Apps > Advanced settings > turn on Developer mode, then Settings > Apps > create a connector with that URL and OAuth, then turn it on in the conversation with the + menu — Infinicia is not in ChatGPT's reviewed app directory yet, so do not tell a person to look for it there. In Cursor: cursor://anysphere.cursor-deeplink/mcp/install?name=infinicia&config=eyJ1cmwiOiAiaHR0cHM6Ly9hcGkuaW5maW5pdGVwZW9wbGUuYWkvdjEvbWNwL2NyZWEifQ== or `{"mcpServers": {"infinicia": {"url": "https://api.infinitepeople.ai/v1/mcp/crea"}}}`. Anything else: the same endpoint. The install page in French, for the person doing it: https://infinicia.ai/fr/mcp/

## Optional

- [Terms of use](https://infinicia.ai/#terms): shown in the app; the service is on invitation during the beta.
- [Privacy policy](https://infinitepeople.ai/privacy): what is processed, for how long, which providers see it, and the rights people have over it.

What a creation costs, to the cent and per plan, next to the published rates of Alibaba Model Studio (Wan), ByteDance (Seedance), Google (Veo, Gemini Image), xAI (Grok Imagine) and ElevenLabs, and next to what ChatGPT, Claude, Gemini and Grok can actually make: https://infinicia.ai/cost/ (English) and https://infinicia.ai/fr/cost/ (French), both with a Markdown twin at index.md. A second of video with sound starts at 0.2 ct, a quick touch-up at 0.8 ct, a minute of speech at 1.4 ct, a three-minute song at 2.9 ct. Computed from the same code that charges the week.

Pricing: https://infinicia.ai/pricing/ (one Infinite People subscription for every trade of the studio: a free first week, then Essentiel 5 EUR, Starter 9.99 EUR, Standard 19.99 EUR — 3x the weekly allowance, Pro 49.99 EUR — 10x and the priority lane).

## Presentations

POST /v1/deck {"prompt", "shape": "deck" | "site" | "quiz", "theme"?, "template"?, "media"?} writes a page with our own model (slides, a one-page site, a live quiz); poll GET /v1/deck/jobs/{id}; result.url is the page, sandboxed, 7 days. POST /v1/deck/jobs/{id}/edit {"instruction"} reworks it in a lineage. GET /v1/deck/kit lists shapes, themes and templates. Full guide: https://infinicia.ai/api/

## Voices

POST /v1/voice/voices keeps a voice on the account: a recording (multipart `file`, `name`, `rights_confirmed`, `transcript` = the text read, for the faithful mode) or a voice designed from words (`name`, `description`, `language`); `sample_url` is the voice to listen to; GET lists them, DELETE forgets one. POST /v1/voice/say {"voice", "text", "seed"?} says a text with it (a creation to poll, WAV 48 kHz, 7 days); POST /v1/voice/convert (multipart `voice`, `file` + `rights_confirmed` | `file_url`) says a recording again in that voice, same words, same pace (WAV 24 kHz). Every plan, GPU seconds on our own machine. Full guide: https://infinicia.ai/api/

## Projects and the bench

A project keeps for good what the studio makes; `GET /v1/projects/{id}` lists its pictures, videos, songs and elements (a voice, a music) and its montages, each with a handle (`picture:5`, `video:12`, `song:3`, `element:7:voice`). Keep a finished video with `POST /v1/projects/{id}/videos {"job_id"}`, a picture with `…/references/from-url {"url"}`, a song or an element with `POST /v1/song/projects/{id}/assets {"job_id"}`, a text said or a recording converted with `POST /v1/voice/projects/{id}/assets {"job_id"}`.

The bench assembles a montage by operations: `POST /v1/projects/{id}/montages {"name"}` opens one in a project, `POST /v1/montages {"name"}` one without a project (the account's inbox keeps what it uses); `POST /v1/montages/{mid}/ops {"ops": [{"op": "add", "track": "video", "from": "video:12"}, {"op": "add", "track": "music", "from": "song:3", "out": 8}, {"op": "transition", "id": "c1a2b3", "kind": "fade"}], "version": 1}` places and changes clips (add, remove, move, trim, split, transition, gain, fade, rename, replace, aspect), all or none; a creation named by its job id (`job:<uuid>`) is kept in the project on the way. Every answer carries `summary` (where each clip sits), `palette` (what can be placed) and `next`. `POST …/montages/{mid}/direct {"prompt"}` does the same from a sentence; `POST …/montages/{mid}/render` exports the mp4 (a creation to poll). Full guide: https://infinicia.ai/api/#montage