Kineo (Kineo AI)
An open-source desktop studio where a multi-agent crew turns a script or novel into 4K anime-style short films.
- Source repo
- Agions/kineo
- Stars
- ★ 137
- Last updated
- 5d ago
- License
- MIT
- Primary language
- TypeScript
- FA score
- 34/100 · Major gaps
At a glance
- Works with
- Portable with changes
- You'll need
- Typical use
- A short-video creator who already has a finished script and wants a voiced, subtitled anime-style draft
- Main limitation
- At Studio v0.0.1 the product is early-stage; behavior and interfaces may still change
- Website
- agions.github.io/kineo ↗ set on GitHub
- Maintenance
- Commits in the last 30 dayslast commit 2026-11-16latest release v0.0.2 (2026-10-21)+0.5 ★/day over 7 days
- Source review
- 34/100 · Major gaps 2 safety controls not found
What does this agent do, and when should you use it?
Kineo is an open-source desktop application built on Tauri v2, React 19 and a Rust engine that takes a script file, a novel chapter, or an AI prompt and drives it toward storyboards and a finished video. Its core is a Hub-and-Spoke multi-agent design: a MasterDirectorAgent coordinates ScriptIngestionAgent, CharacterDesignerAgent, StoryboardArtistAgent, SoundEngineerAgent and VideoEditorAgent around a shared ProjectBlackboard, with user-defined agents that can be registered and pinned to a trigger stage. The project claims integration with 13 AI providers across text LLMs, image generation and 4K video engines, plus GPU-accelerated encoding through Apple VideoToolbox or NVIDIA NVENC. API keys are stored in the OS-native Keyring and project data stays local. The current release is Studio v0.0.1 (Multi-Agent Edition), so it is early-stage software.
After you import a .txt/.md/.docx/.json script, novel text or prompt, ScriptIngestionAgent extracts the chapter outline and scenes. CharacterDesignerAgent pulls out character entities and produces persona prompts plus consistency anchors. StoryboardArtistAgent plans Zoom/Tilt camera moves and animation prompts and lays out a three-column storyboard board. SoundEngineerAgent calls EdgeTTS or CosyVoice for multi-track voice and computes subtitle timing. VideoEditorAgent schedules the render queue and exports 4K video using Apple VideoToolbox (Metal GPU) or NVIDIA NVENC. MasterDirectorAgent watches ProjectBlackboard state and dispatches work to child agents, while any CustomUserAgent you add in the UI joins the AgentRegistry and the blackboard pipeline at stages such as after-script, after-storyboard or after-audio. The derived result can be loaded straight into the Step 3 storyboard editor without re-uploading material.
- A short-video creator who already has a finished script and wants a voiced, subtitled anime-style draft
- A web-novel author turning a chapter into consistent character art and a storyboard trailer
- A director prototyping several camera-move storyboards from one prompt before production
- An indie animation team inserting its own review or rewrite logic as a custom agent after script parsing
- A creator who wants project files and API keys kept on the local machine rather than in a cloud account
How do you install or deploy this agent?
The repository does not document an official prebuilt installer; the README path is building from source. Prepare Node.js, pnpm, a Rust toolchain and the Tauri v2 system prerequisites first.
How do you use this agent?
On first launch, configure API keys for the AI providers you intend to use (they are stored in the OS-native Keyring). Then open the multi-agent creation center and import your material to start the agent pipeline.
What are this agent's strengths and limitations?
- The Hub-and-Spoke plus ProjectBlackboard split keeps each child agent scoped to one job, so a single stage can be swapped or extended
- Custom agents can be declared in the UI with a trigger stage (after script, after storyboard, after audio), without touching the orchestration core
- One pipeline covers text parsing, character consistency, storyboard planning, TTS voice and GPU 4K encoding, and its output feeds directly into the storyboard editor
- API keys live in the OS-native Keyring and project data stays on disk, lowering exposure of unpublished material
- MIT-licensed, so teams can fork and redistribute
- At Studio v0.0.1 the product is early-stage; behavior and interfaces may still change
- Core generation requires third-party AI provider API keys, so real usage incurs metered costs
- Building from source needs a Rust toolchain and Tauri v2 system dependencies, which raises the entry cost
- GPU 4K encoding relies on Apple Silicon VideoToolbox or NVIDIA NVENC, so export speed depends on your hardware
- Several model names and release dates in the README point to future or unverifiable versions, so those integrations are hard for outsiders to confirm
How does this agent compare with similar options?
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| Kineo (Kineo AI) This agent | 34 · Major gaps | — | ★ 137 | 5d ago | TypeScript | — |
| OpenHuman | 57 · Major gaps | Desktop appFreemium | ★ 43k | 1d ago | Rust | — |
| OpenMAIC Multi-Agent Interactive Classroom | 79 · Good | Web appFree + model costs | ★ 44k | today | TypeScript | — |
| Due Diligence Agents | 69 · Some gaps | CLIFree + model costs | ★ 139 | 3d ago | Python | Claude API |
How does FollowAgents rate this agent?
Why each dimension lost points
Evidence shows claims of Keyring-encrypted API keys and local-only project data, but the configuration doc also offers plaintext .env variables; no permission declarations, confirmation flows, or rollback exist, and plugin hooks can rewrite scripts/characters without user consent. Deductions: no least-privilege statement, no user confirmation, no rollback/undo, no dependency audit evidence; scored 1 only on README assertions.
Code has try/catch and logger error output in PluginRegistry (thin failure messaging), but README claims 13 providers while configuration doc lists 7, and model names (DeepSeek-V4, Qwen 3.8-Max, GPT-5.6) are unverifiable, undermining self-consistency. Deductions: doc/code mismatch, no lockfile evidence for dependency availability, failure info limited to logs.
README and installation doc clearly define target users (comic-drama creators, short-video bloggers, directors) and three input scenarios, with explicit cross-platform system requirements; but capability boundaries (what is truly automated vs Mock) are vague and trigger stages are only exemplified. Deductions: weak boundaries and trigger precision.
Docs structure (getting-started, user-guide links, VitePress) and install steps are complete, and the MIT LICENSE is full and standard; however version numbers conflict (README v0.0.1, package.json 0.0.2, doc frontmatter >=3.0/>=2.2), there is no CHANGELOG, no known-limitations section, and maintenance responsibility is only attributed to Agions with an unverified publisher. Deductions: version inconsistency, missing changelog and limitations.
Product positioning is clear and output target (4K comic-drama video) is concrete, but samples contain only utility functions and a plugin registry, with no actual generation pipeline code or output samples, so output usability cannot be judged; the claimed 40-70% speedup and 13-model matrix lack verifiable support, making marginal value and cost-benefit hard to assess. Deductions: no output examples, no cost data.
README heavily uses future-year (2026) and unreleased model names, which are untraceable marketing claims; docs and code contradict each other on provider count and version numbers, and facts are not separated from inference. Deductions: untraceable claims, cross-source inconsistency, no fact/inference separation.
- Not found in source: confirmation before actingTurn on (or add) a confirmation step before it acts, and try it in a sandbox or test environment before real data.
- Not found in source: rollback or recovery pathBack up first, or work on a git branch or snapshot, so its changes can be undone.
- README claims 13 providers while the configuration doc lists 7, and model names (DeepSeek-V4, GPT-5.6) are unverifiable; do not assess capability from these claims.
- Plugin hooks can rewrite scripts and character data without confirmation and there is no rollback; review third-party plugins before use.
- The configuration doc offers plaintext .env API keys, in tension with the README's Keyring encryption claim; watch for key leakage.
- Version numbers conflict across README, package.json, and doc frontmatter, with no CHANGELOG; verify the actual version before upgrading.
- Publisher identity is unverified; maintenance responsibility and update path are unclear.
Recent releases
Latest entries from the project's GitHub Releases. Notes are quoted from upstream as written, not rewritten or tested by FollowAgents.
-
Release notes excerpt
Download Kineo (Kineo AI) 0.0.2 | Platform | Architecture | Download | |----------|-------------|----------| | Windows | x64 | [.exe](kineo-0.0.2-windows-x64.exe) | | Windows | ARM64 | [.exe](kineo-0.0.2-windows-arm64.exe) | | macOS | Universal | [.tar.gz](kineo-0.0.2-macos-universal.tar.gz) | | Linux | x64 | [.tar.gz](kineo-0.0.2-linux-x64.tar.gz) | > Full release notes: [CHANGELOG.md](https://github.com/Agions/kineo/blob/v0.0.2/CHANGELOG.md)View on GitHub ↗ -
Release notes excerpt
Download Kineo (Kineo AI) 0.0.1 | Platform | Architecture | Download | |----------|-------------|----------| | Windows | x64 | [.exe](kineo-0.0.1-windows-x64.exe) | | Windows | ARM64 | [.exe](kineo-0.0.1-windows-arm64.exe) | | macOS | Universal | [.tar.gz](kineo-0.0.1-macos-universal.tar.gz) | | Linux | x64 | [.tar.gz](kineo-0.0.1-linux-x64.tar.gz) | > Full release notes: [CHANGELOG.md](https://github.com/Agions/kineo/blob/v0.0.1/CHANGELOG.md)View on GitHub ↗
FAQ
Do I have to pay to use Kineo?
Are my assets and API keys uploaded anywhere?
Can I run it without a discrete GPU?
What can a custom agent actually do?
Is Kineo (Kineo AI) open source?
Is Kineo (Kineo AI) safe to use?
Is Kineo (Kineo AI) still maintained?
Who makes Kineo (Kineo AI)?
What is Kineo (Kineo AI)'s official website?
Thanks — we use this to improve the page.