Getting Started with the AIFISHER Canvas: Connecting Assets, AI Models, and Creative Workflows

Introducing the AIFISHER Canvas

AIFISHER is a canvas tool for AI creation that brings images, video, audio, text, and AI workflows into one space. You can organize ideas and assets on the canvas, connect model services, plan storyboards, and save generated results, turning scattered operations into a traceable workflow.

The project supports connections to the AIFISHER API, other model services, local ComfyUI, and RunningHub. It also provides Agent-assisted planning and operations, API workflow import, parameter configuration, and more. Projects and assets are stored locally by default, but when you call online models, prompts and reference assets are sent to the service provider you select.

Core Features

• Multimedia nodes: Connect image, video, audio, and text nodes on the same canvas.
• Model service integration: Use the AIFISHER API or configure other providers, local ComfyUI, or RunningHub.
• Reference and generation: Reference existing assets, adjust generation parameters, and connect the results to subsequent workflows.
• Creative workflow management: Use Agent-assisted planning and operations to keep ideas, storyboards, prompts, and results in the project.
• Local project storage: Projects and assets are stored locally by default. In the source version, the workspace is located in .desktop-dev/ under the runtime directory.

Installation and Runtime Requirements

Running from source requires the following:

• Windows x64
• Node.js 24.20.0
• npm
• ComfyUI must also be installed if local generation is required
• A valid API Key from the relevant provider is required to use online models

Video transcoding and audio processing also require working FFmpeg and ffprobe. The source snapshot does not include these runtime binaries and does not automatically install ComfyUI or model weights.

Starting AIFISHER from Source

First, obtain the project source code through Git and open PowerShell in the project root. Then install the dependencies and start desktop development mode:

npm ci
npm --prefix apps/desktop ci
npm run desktop:dev

On the first launch, the Electron desktop runtime downloads using the checksum included with the npm package, so a network connection is required. After startup, the source version's local workspace is stored in the .desktop-dev/ directory. This directory contains local work data and should not be committed to the repository or shared with others.

Verifying Project Status

After installation, you can run the project check command:

npm run check

This command performs type checking, tests within the public scope, and builds the frontend and desktop launch pages. Automated tests do not call paid generation services, so the check process does not incur model usage fees.

Configuring the AIFISHER API

Online model services require separate registration and configuration. Follow these steps:

1. Open AIFISHER API and register or sign in.
2. Create and copy your own API Key on the API site.
3. Launch the AIFISHER Canvas and open Settings.
4. Open Closed-Source Services and select AIFISHER API.
5. Paste the API Key and save it.
6. Return to the canvas and select the appropriate online model service as needed.

The canvas account and API site account are managed separately. The local canvas can be used without signing in; online generation is billed according to the provider's actual policies, and the software does not include free model credits. It is recommended that you save keys only through the Settings interface and never write real API Keys into example configurations or commit them to a Git repository.

Organizing a Creative Session on the Canvas

A basic workflow is to first import or create text, image, and other asset nodes, then connect model services according to your creative goals, and finally save the generated results. You can place prompts, reference images, and parameters in the same workflow for easier iteration.

1. Define the subject: Use text nodes to record creative goals, character settings, scene descriptions, or storyboard information.
2. Add reference assets: Connect image, video, or audio nodes as references for visual, motion, or sound direction.
3. Connect a model: Select the configured AIFISHER API, another online service, ComfyUI, or RunningHub.
4. Adjust parameters: Modify generation parameters according to the options provided by the model and service provider.
5. Review the results: Keep generated results on the canvas and compare them with the original prompts and reference assets.
6. Continue iterating: Modify prompts, replace reference assets, or adjust parameters to generate new versions.

The project includes style and camera movement asset libraries with the canvas for exploring visual directions. For example, the libraries include animation styles, lifestyle portraits, and science-fiction visuals, as well as camera movement references such as orbiting ascent and the Hitchcock zoom. Actual generation results still depend on the model, prompts, and reference assets used.

Connecting Local and Third-Party Workflows

Using Local ComfyUI

To connect local ComfyUI, first install and run ComfyUI separately on your system, then configure the corresponding connection in AIFISHER. When reporting ComfyUI issues, include the smallest reproducible API JSON to help identify workflow or parameter problems.

Using RunningHub or Other Providers

AIFISHER also supports RunningHub and other model services. Before using these services, prepare a valid API Key for each one and configure it according to the provider's pricing, privacy, and usage conditions. API workflows can be imported and configured further, but the specific capabilities available depend on the connected service.

Data, Privacy, and Diagnostic Considerations

AIFISHER stores projects locally by default, but a “local canvas” does not mean that every model runs locally or that assets will never be uploaded. When you actively call an online model, the relevant prompts and reference assets are sent to the selected provider. Review its pricing, privacy, and content usage policies first.

Official account and device identity services connect to the network. Even without signing in, device identifiers and diagnostic information may be sent to official services. Diagnostic information includes the application version, model and provider, call status, duration, error stage, HTTP status, request or task identifiers, and a desensitized error summary. Diagnostic reporting is not intended to collect API Keys, complete prompts, or original assets. Files you actively attach when submitting feedback are sent to the feedback service.

Advanced Usage Recommendations

• Protect your keys: Always save API Keys through the Settings interface. Avoid putting real keys in configurations, logs, or example code.
• Manage the workspace: Do not commit or share .desktop-dev/, which may contain local work data.
• Start with small-scale verification: Use npm run check to verify the source environment. Automated tests do not call paid generation services.
• Distinguish local and online models: Before sending reference assets, confirm the connected model service and the scope of data transmission.
• Prepare reproducible feedback: When reporting an issue, provide the version, reproduction steps, and desensitized logs. Remove accounts, API Keys, private paths, and private assets.
• Handle media dependencies: When video transcoding or audio processing is involved, confirm that FFmpeg and ffprobe are available.

Installer Version and License

If you do not need to run from source, visit the release page or access the official download drive to obtain the installer version. Installed official releases will continue to receive signed automatic updates.

The project source code is limited to non-commercial use. Non-commercial learning, research, modification, and free redistribution are permitted, but commissioned work, commercial production, resale of installation packages, commercial derivative development, and paid hosting are prohibited. The project uses the AIFISHER Noncommercial Source License 1.0 and is not an open-source license as defined by OSI. Third-party components follow their original licenses. Closed-source SKILL, private shared templates, personal accounts, and keys are outside the public scope.

Before using AIFISHER, read LICENSE, SOURCE-SCOPE.md, and THIRD-PARTY-NOTICES.md in the repository and confirm that your use complies with the license terms.

Summary

The AIFISHER Canvas is suited to managing multimedia assets, prompts, model services, and AI creative steps in one place. To get started, deploy the source or installer version on Windows, configure an API Key through Settings, and then build a workflow starting with text and reference asset nodes. When using online models, pay particular attention to costs, privacy, key protection, and the boundaries of the non-commercial license.