Refly

Still switching between multiple tools for content creation? Want to do graphic, audio, video linkage content, but the technical threshold is stuck? Refly’s emergence has reconfigured the logic of AI creation – as an open source AI native content creation platform, it is based on the “free canvas” as the core, integrating multi-modal generation, knowledge base integration, code visualization and other capabilities, so that creators do not need to switch tools to complete the entire process from creative dissemination to finished output within a single interface. As an open source AI native content creation platform, it takes “free canvas” as its core, integrating multi-modal generation, knowledge base integration, code visualization and other capabilities, so that creators don’t need to switch tools, and complete the whole process from creative dissemination to finished product output within one interface. This article combines the latest v0.9.0 version of 2025 with the actual test to dismantle its core functions, operational scenarios and unique advantages, helping users with different needs to quickly unlock a new way of efficient creation.

Refly’s core positioning: the “all-round creative workstation” in the AI era.

Refly’s core mission is to “let creators focus on thinking, AI help efficient landing”, it is not a single-function tool, but a full-stack platform that integrates multimodal creation, knowledge management, workflow scheduling, and code generation. Its positioning can be summarized as “a creation tool for three types of users”:

• Content creators: generate graphics, audio, and video with one click to realize multi-platform content linkage;
• Knowledge workers: Integrate literature and webpage materials, intelligently organize the research framework, and improve the efficiency of report and thesis writing;
• Developers/designers: quickly generate code prototypes, design sketches, and support custom models and plug-in extensions.

Different from Dify’s enterprise-level application orientation and Trickle AI’s zero-code website building, Refly’s biggest advantage is its “full coverage of creation scenarios” – not only retaining the low threshold of visualization, but also opening up open-source deployment, custom models, and other advanced features, especially multimodal modeling. Refly’s biggest advantage is its “full coverage of creation scenarios” – it not only retains the low threshold of visualization, but also opens up open-source deployment, custom models and other advanced features, and especially excels in multimodal content linkage and creative dispersion scenarios. As of November 2025, its GitHub repository has gained 2,000+ starred labels, and the v0.9.0 version has a credit-based billing system and a brand-new interface design, which further improves the flexibility of use and smoothness of experience.

Second, the core function of the actual test: 5 highlights, redefine the AI creation

1. Free Canvas + Multi-threaded Dialog: Creative Diffusion without Boundaries

This is Refly’s signature feature. The canvas interface supports parallel creation of multiple topics, and the multi-threaded dialog allows visualization of thoughts. The steps to build a “product promotion content package” are as follows:

1. Create a new project and enter the canvas, drag and drop the nodes of “Text Generation”, “Image Generation” and “Audio Dubbing” on the left side to build the content production chain;
2. Initiate a multi-threaded dialog for product selling points: one thread generates Little Red Book copy, one thread refines the core keywords, and the threads can quote each other’s content without copying and pasting over and over again;
3. Click on the correlation line between the nodes, set logic rules (such as “automatically trigger image creation after text generation”) to automate the process.

The canvas supports node dragging and dropping, batch editing, and can also switch between light and dark themes to fit a long period of time to create, knowledge workers use it to sort out the framework of the paper, you can simultaneously parallel “literature excerpts”, “rebuttal of ideas”, “data organization When knowledge workers use it to sort out the framework of the paper, they can simultaneously parallel “literature excerpts”, “viewpoint rebuttal”, “data organization” and multiple threads, which will increase efficiency by more than 50%.

2. Multi-modal full-link creation: a text generates graphics, audio and video.

Refly v0.8.0 has unlocked the ability to generate images, audio and video, and v0.9.0 further optimizes the multimodal linkage, and it only takes 15 minutes to test the whole process of “short video scripting → dubbing → screen generation”:

1. Input the short video script into the canvas, call the “Text to Audio” node, select the Lyria-2 model to generate the background music, and complete the AI dubbing with the Chatterbox model, which supports adjusting the speech speed and emotion.
2. Associate the voice-over file to the “Video Generation” node, enter the description of the screen “technological product display, dynamic particle effect”, and select the SeeDance-1-Lite model to quickly generate short video clips; for optimization, directly use the “Text Generation Audio” node to generate the background music, with the Chatterbox model to complete the AI voice-over, supporting the adjustment of speech speed and emotion.
3. If you need to optimize, directly modify the text command on the canvas, AI real-time update of the corresponding audio or video content, no need to jump to a third-party tool.

Its multimodal advantage is to support the “generation → analysis → re-creation” closed loop, for example, after generating product posters with the Flux-Pro model, the image analysis node can be called directly to extract visual keywords, which can then be used to optimize the copy to ensure that the content style is uniform. As shown in Figure 2, the multimodal nodes support drag-and-drop combinations, so that non-technical personnel can also quickly build a complex creation process.

3. Credit-based billing + dual-model configuration: flexible adaptation to different needs

The core update of v0.9.0 is the introduction of a credit-based billing system, which completely simplifies the process of using models:

• No need to manually configure API keys, all models (text/image/audio/video) are billed by credit consumption, which is transparent and verifiable;
• Free users receive basic credits every month, and early unlimited members can continue to use GPT-4.1, Kimi and other mainstream models for free;.
• Support dual model configuration mode: global mode automatically loads recommended models, zero configuration to get started; custom mode can add DeepSeek, Gemini and other third-party models to retain personalized settings.

The test found that generating a high-definition product image only consumes 5 credits, and a 1-minute AI voiceover consumes 3 credits, which is more cost-effective than similar multimodal tools, so individual creators do not need to worry about the high cost of use.

4. Knowledge Base + Intelligent Acquisition: Material Integration without Worries

Refly has a powerful built-in knowledge base engine that supports the import of PDF, Word, web pages and other 7+ formats, and with the Chrome plug-in, you can cut and hide the contents of GitHub, Medium and other platforms with a single click. Test the “academic paper writing” scenario:

1. Through the plug-in to cut and hide 3 related documents, the knowledge base automatically complete the text chunking and semantic indexing;
2. Call the “Knowledge Base Search” node in the canvas, enter the keyword “AI multimodal development trend”, AI quickly extract the core viewpoints of the three pieces of literature and label the source;
3. Associate with the “Text Generation” node to automatically generate a literature review framework based on the search results, and support one-key export to Markdown format.

The knowledge base supports the visualization of knowledge maps, which can automatically associate related materials, helping creators to explore cross-field inspiration and avoid homogenization of content.

5. Code generation + real-time preview: an efficient tool for developers and designers.

Refly has a built-in code generation engine that supports HTML, SVG, Mermaid and other formats, and real-time preview. Test “quickly build product prototype page”:

1. Input the requirement of “responsive product page with rotating charts and forms” into the canvas;
2. Call the “Code Generation” node, select the React framework, AI seconds to generate the complete code.
3. Click the preview button to view the page effect on the right side of the canvas, modify the code directly or adjust it through natural language commands (e.g., “change the color of the button to blue”), and synchronize the update in real time.

This feature is extremely friendly to non-professional developers, a self-media blogger used it to generate SVG data charts, and then with the multimodal function to generate the explanation video, the whole process does not need to rely on the technical team, the content output cycle from 3 days to half a day.

Typical application scenarios: covering the whole process of creation

1. Self-media multi-platform content linkage

Creators can generate public website copy, Xiaohongshu graphic, short video script + dubbing + screen at one time in Refly, associating the core elements of different platforms’ contents through canvas nodes to ensure a unified style with different focuses. After a beauty blogger used Refly, the efficiency of multi-platform content output was increased by 300%, and the fan growth rate was accelerated by 2 times compared with the previous one.

2. Academic and workplace report writing

When students write a paper, they can integrate the literature through the knowledge base, use multi-threaded dialogues to sort out the arguments and data, and AI automatically generates a structured framework and labels the citation sources; when people in the workplace do a work report, they can import Excel data with one key to generate visual charts and summary text, saving 80% of the typesetting time.

3. Product Design and Promotion

Designers can quickly generate product sketches and KV posters, developers can synchronize the generation of page prototypes, marketers can create promotional texts and short videos based on the design materials, and the whole team can collaborate on the same canvas, avoiding repeated communication and confirmation, and shortening the product on-line cycle by 40%.

4. Corporate training and internal communication

HR department can set up a “training content creation flow”: generate training documents, supporting audio lectures, case videos, and store them in the knowledge base; when employees query, AI can intelligently answer their questions based on the content of the knowledge base, which reduces the cost of training and the threshold of communication.

Comparison with similar tools: Refly’s core competitiveness

Data source: Test experience and official document collation

V. Precautions for Use and Guidelines for Avoiding Pitfalls

1. Environmental requirements: Chrome 110+ or Edge 100+ is recommended to avoid problems such as node dragging and dropping failures and preview anomalies in older versions.
2. Deployment Configuration: Private deployment needs to meet the 2-core 4GB memory, support Docker Compose one-click start, visit localhost: 5700 after deployment can be used.
3. Credit consumption: free version of the monthly credit is limited, high-frequency use of multimodal features (such as video generation) is recommended to upgrade the Pro version, more cost-effective;;.
4. Model Selection: Flux-Schnell model is recommended for generating fast sketches, Flux-Pro is recommended for high-definition commercial sketches, and video generation is preferred to SeeDance-1-Lite (fast) or SeeDance-1-Pro (excellent image quality).
5. Data security: private deployment is suitable for storing sensitive materials, cloud version is recommended to avoid uploading confidential files, and custom models need to keep the API key properly.

Conclusion: Creativity without boundaries, efficient creation.

With the trinity structure of “free canvas + multimodal + knowledge base”, Refly breaks down the barriers between creative tools, allowing users from different backgrounds to complete the whole process from inspiration to finished product output in one-stop. Whether it’s self-media content creation, academic research, product design and corporate training, the flexible combination of features can improve efficiency, focusing on the core creativity rather than the operation of the tool.

As a continuous iteration of the open source platform, Refly will launch desktop clients, offline functions and plug-in market to further expand the use of scenarios and flexibility. Visit the official website ( https://refly.ai ) to sign up for free, or deploy a private version through the GitHub repository ( https://github.com/refly-ai/refly ) to start your multimodal AI creation journey.