AVA STUDIO

In the first version, users were greeted by a large prompt box as the main entry point. They could directly enter a video generation prompt to create a video. At this stage, Ava Studio functioned more like a general video creation tool: users generated a video, moved into the editor for further adjustments, and then exported the final result.

In the second version, we introduced the concept of “Studio” alongside the existing editor. Instead of relying on one large prompt box on the homepage, users could start their creation process through more specific entry points based on different needs.

For example, when users selected “Clone Viral Video,” they would upload a reference video. The agent inside Studio would then analyze the uploaded file, break it down into different scenes, and help recreate the video structure.

In the third version, we introduced the concept of “Campaign.” At this stage, the product shifted away from serving general video creators and became more focused on marketing-related creative teams.

The homepage was redesigned around campaign workflows, helping users quickly launch different types of campaigns, validate creative directions through research, discover high-performing templates, and clone successful short-form social media videos for both organic content and paid ads.

In the first version, the onboarding flow focused on extracting information from the website or materials provided by the user. Ava Studio would generate a brand profile, and users were asked to review, confirm, and edit the Brand DNA before receiving text-based creative ideation.

After that, users needed to enter Studio to further generate and edit their videos. This flow helped users build a structured brand foundation, but the output was still mostly text-based, which made the value of the product less immediately visible.

After testing the first version, we found that users had difficulty understanding the final video outcome from text-based ideation alone. Since they could not immediately see what the generated creative would look like, the onboarding experience felt less engaging, and we observed a higher drop-off rate at this stage.

In the second version, we redesigned the flow to bring the aha moment forward. Instead of asking users to edit the brand profile first, Ava Studio directly analyzed the website or uploaded materials and recommended video templates that matched the user’s brand and marketing needs.

Users could then select a template and enter the editor directly, where they could quickly swap the cast, text, and other key elements. This made the experience more visual, more actionable, and easier to understand, helping users see the product’s value much earlier in the journey.

Overall, the onboarding evolved from a brand setup flow into a result-driven creation flow. By reducing upfront configuration and showing users tangible video templates earlier, we shortened the path to value and improved the likelihood of users continuing into the editing experience.

In the first version, we introduced the Studio workflow, along with the concepts of “@ingredients” and “scenes.”

Inside Studio, users could add scenes from the left-side scene thumbnail panel. Each scene worked as an independent generation space. When users were satisfied with a generated video or image, they could set it as the final scene. Only the selected final scenes would appear in the Editor for further editing and export.

Within the Studio workspace, users could add assets as ingredients and directly reference them in the central prompt box through @mentions during video generation.

At this stage, the Feed View was organized around prompt requests. Each prompt was strongly connected to its corresponding generations, making it easy for users to understand which prompt produced which results. The benefit of this structure was clarity: users could clearly see the relationship between each prompt and its generated outputs. However, the downside was that when a prompt only produced one generation, the prompt-generation group still took up a large amount of workspace, making the overall layout less space-efficient.

In the second version, we introduced agents into Studio. Based on the user’s goal, the agent could create a generation plan and automatically write prompts for each scene. Users could then generate videos for all scenes directly through the agent with a single action.

In terms of layout, we moved away from the rigid prompt-generation grouping and adopted a more flexible, responsive layout. This helped save workspace space while allowing users to iterate on content more efficiently.

In the latest version, we introduced the Node Editor to make the generation and iteration process more visual and intuitive.

Studio is now divided into two generation modes: the original Feed View and the new Node Editor. In the Node Editor, users can freely drag from ports to create new nodes, making it easier to visualize the full iteration process behind each generation.

If users are not satisfied with a result, they can quickly trace back to the specific node where the issue occurred, adjust the input or prompt, and regenerate from that point. This makes the generation workflow more transparent, controllable, and efficient.

In the first version, the editor was designed to support basic video editing needs as well as AI chat-based video generation. Users could select individual clips, regenerate them through chat, and adjust basic settings such as timing, visuals, and other clip-level properties.

In the second version, we introduced the concept of “Assets” while keeping the existing settings and chat-based generation features. Users could upload or generate images, videos, audio, and other media assets, add them to the timeline, and use these assets as the foundation for further video generation.

This made the editor more flexible, shifting it from a simple clip-editing tool into a more asset-driven creation environment.

In the third version, the editor started to work more closely with Studio. Studio became the main stage for video generation, while the editor focused more specifically on video refinement, timeline adjustments, and final edits.

At this stage, we also introduced the concept of “@ingredients,” which will be explained in more detail in the Studio section. An ingredient can be understood as a reusable variable within each project. Users can call an ingredient through an @mention during generation. When a user replaces a mentioned ingredient, all related clips can be updated accordingly, significantly improving editing efficiency and reducing repetitive manual work.