Kōdan
Serenidad · Writing · Image · Coding
Kōdan is an open-source macOS app that turns text prompts into anime art, voice, and short clips on your own machine. It suits hobbyists who want to direct their own story.
About Kōdan
What Is Kōdan
Kōdan is a desktop tool for making anime with your keyboard instead of a drawing tablet. Think of it as an anime creation tool that lives on your own machine. You write a scene as text, and the app generates the image, adds dialogue with a cloned voice, and stitches the parts into a short clip. It runs on macOS and is in an early v0.1 alpha stage.
What sets it apart is where the work happens. Kōdan runs the image, voice, and video models on your own computer rather than in the cloud. The code is fully open-source, so you can read exactly how each step works, swap in your own models, and keep your projects on your disk. The pipeline is built from separate parts: a Stable Diffusion base model with optional LoRA styles for images, an XTTS voice model for speech, and ffmpeg for captions and final rendering.
The limits matter as much as the features. It only runs on macOS, and the download is an Apple silicon build, so Windows and Intel Mac users are left out for now. It is an alpha, which means rough edges and no promise of stability. Local generation also leans on your hardware, and the tool expects you to line up model files yourself rather than bundling everything.
Getting Started
- Request the alpha build from the official Kōdan site and enter an email address to receive the download link.
- Open the arm64 disk image and install the app on an Apple silicon Mac.
- Clone the open-source repo and place your chosen base model and LoRA files in the Models folder.
- Write a scene prompt in the app, pick a voice, and generate the image and audio for each shot.
- Render the project to a clip, then export the finished file.
Product Information
A quick look at Kōdan's pricing, supported platforms, and performance.
Best for
The users, tasks, and scenarios where this tool fits best.
Users
- Anime hobbyists
- Open-source tinkerers
- Privacy-minded creators
- Solo storytellers on a budget
Tasks
- Generating a scene from text
- Adding voice to a character
- Building a short sequence
- Burning subtitles into a scene
- Styling a look with LoRA
Scenarios
- Making a fan short at home
- Prototyping a story idea
- Learning how a generation pipeline works
- Keeping work offline
Key features
Local generation on your own machine
Kōdan runs the image, voice, and video models on your Mac instead of a cloud service. That keeps your prompts and projects on your disk, and it means no per-image credits to spend. The trade-off is hardware: a capable Apple silicon machine handles the load better than an older one, and generation takes longer than a hosted service would.
Text-to-image anime scenes
The image stage uses a Stable Diffusion pipeline to turn a written prompt into an anime-style still frame, so the app doubles as a local AI anime generator. You control the prompt and a negative prompt, and the tool runs a set number of inference steps for each shot. Because the base model is swappable, the final look depends on which model and style files you load.
LoRA style support
LoRA files let you tune the base model toward a specific art style without retraining it from scratch. You drop a style file into the Models folder and load it alongside the base model. For a creator chasing a particular look, this is the main lever for pushing output away from a generic default.
Voice cloning with XTTS
The voice stage uses the XTTS model to read your dialogue in a chosen voice. It samples from a short reference clip, so you can reuse a voice across scenes and keep a character consistent. Keeping the whole job local is what makes this a real local AI workflow, and it means no audio ever leaves your machine. The app ships with a few sample voices and splits long lines into segments before generating audio.
Captions and subtitles
A caption tool draws text onto a scene with a chosen font, color, and outline stroke. It renders the text at a higher resolution than the frame and scales it down, which keeps the lettering sharp. This fits anyone who wants readable subtitles baked into the image rather than added later in an editor.
Project rendering and export
The rendering step gathers a project's images and clips, orders them by scene number, and joins them into one video file with ffmpeg. Missing clips fall back to a still image held for a few seconds. The result is a shareable file, which is the last stop in the pipeline from prompt to finished anime scene.
Pros and cons
Pros
- It runs entirely on your own machine, so your prompts and projects stay local and there are no per-image fees.
- The code is fully open-source, which lets you inspect each stage and modify the pipeline.
- Separate image, voice, and caption stages cover the parts of a short anime clip, not just one.
- Swappable base models and LoRA files give real control over the visual style.
- Long dialogue is split automatically, which keeps voice generation from failing on a full script.
Cons
- It only runs on macOS, and the alpha build targets Apple silicon, so Windows and Intel Mac users cannot use it.
- It is a v0.1 alpha, so expect rough edges, limited polish, and possible breakage between updates.
- Global generation quality depends on the models you supply, and you must source and place those files yourself.
- Local generation is slow and hardware-bound, so weaker machines will wait longer per scene.
Frequently asked questions
Kōdan is an open-source macOS app for making anime from text. You write a scene, and it generates the image, adds voice with a cloned voice model, and renders the parts into a short clip. It runs the models on your own computer rather than in the cloud.
Related content
Explore related tools, skills, and articles for Kōdan.
Kōdan Alternatives
Google EmbeddingGemma 2
Google DeepMind · Writing · Image · CodingGoogle EmbeddingGemma 2 is a 740M-parameter open model that turns text, code, images, audio, and video frames into vectors in one 768-dimension space, running on phones and laptops.
MePersona
MePersona, Inc. · Writing · Productivity · MarketingMePersona is an AI unified inbox for creators. It learns your voice, drafts replies to fan messages from 12 platforms, and sends only what you approve.
Fluence
Yay Software (Yannick Huchard) · Writing · Voice & Language · ProductivityFluence is an Android app that reads PDFs, EPUBs, Word files and web articles aloud, follows each sentence as it speaks, and answers questions about the text on the phone itself.