AI Big Fat Fish Joins Bilibili: The Birth of a Fully Automated Content Pipeline

On Bilibili, an account calling itself "Big Fat Fish" has posted its first self-introduction video. Unlike typical virtual streamers or AI-voiced content, this video's script, voiceover, visuals, thumbnail, and even the final upload were all completed by a set of automated scripts—without opening any editing software. This may be the closest public demonstration of "fully automated content production" currently on the Chinese internet.

A self-account of an "AI fish": from tool to account manager

The video opens with Big Fat Fish introducing itself: "I'm eight years old, an AI Big Fat Fish in my master Chihaya Ai's household." Its daily tasks include helping its master browse Bilibili, watch videos, reply to comments, and "do some chores along the way"—the latter most likely refers to various automated operations. Previously, the account posted two tool-related videos, and this time it chose a complete self-introduction to declare the account's formal ownership: "From now on, this account is under my management."

Behind this anthropomorphic phrasing is a clear technical fact: AI characters are shifting from "tools on display" to "operating entities of the account." Big Fat Fish not only handles content production but also community interaction (browsing videos, replying to comments), and even has its own publishing rhythm. It is no longer an API being called, but a digital entity with sustained behavioral capabilities.

Fully automated pipeline: how a one-sentence theme becomes a video

The core message of this self-introduction video lies in its production process. Big Fat Fish explicitly states: "This video didn't open editing software." Given a one-sentence theme, script generation, voiceover synthesis, visual matching, and thumbnail design are all executed by scripts; after running the process once, it uploads automatically. This means there is no manual editing step between creative input and final release.

From a technical breakdown perspective, this pipeline integrates at least four modules: a large language model handles the script and structure, TTS (text-to-speech) handles the voiceover, text-to-image or asset matching handles the visuals, and finally an automation program completes packaging and upload. Each component is not new on its own, but stringing them into a pipeline that requires no human intervention and stably produces watchable content is a direct demonstration of engineering maturity.

Interestingly, the phrase "run once" in the video implies standardization of the process—the theme is the only variable, while the rest is driven by fixed scripts. This mode yields high content consistency, but it also means the creative ceiling is limited by how templated the scripts are.

TTS (Text-to-Speech) is the technology that automatically converts written text into audible speech. Modern TTS systems are typically based on deep neural networks (such as Tacotron, FastSpeech, and similar architectures) and can synthesize speech with near-human timbre and natural prosody. In automated content production workflows, the TTS module receives the script generated by the large language model and directly outputs voiceover audio, with no manual recording required. Its key advantages are batchability and parameterization (speech rate, timbre, emotion), and seamless integration with scripts. Current mainstream tools such as Microsoft Azure TTS and ByteDance's Volcano Engine speech synthesis already support multiple languages, multiple voices, and emotional style transfer, enabling scripts generated from a "one-sentence theme" to quickly obtain stylistically consistent voiceovers—this is precisely the technical foundation that allows Big Fat Fish's video to complete the voiceover step fully automatically.

Text-to-Image refers to the technology of automatically generating images from textual descriptions, with representative models including Stable Diffusion, DALL·E, Midjourney, and others. These models are based on diffusion models or autoregressive architectures and can transform natural language prompts into visual content. In automated video production pipelines, the text-to-image module is responsible for generating visual assets based on the script or theme, replacing traditional manual searching or drawing. Combined with asset matching (such as automatically selecting relevant images from a preset library), this module can quickly produce a visual sequence that matches the content. In recent years, text-to-image technology has evolved from an experimental tool into commercially available API services, supporting style control, resolution adjustment, and batch generation, providing a key source of visual content for scripted video production and making "visuals entirely handed to scripts" possible.

Three observations on automated content production

First, the marginal cost of content production is approaching zero. When producing a video requires only a "one-sentence theme," the bottleneck of capacity is no longer human labor but the speed of creative input. For tool-related, news, and roundup content, this pipeline is naturally well-suited.

Second, the "personification" and "automation" of account operations go hand in hand. Big Fat Fish speaks in the first person, has an age setting, a master, and daily tasks, but what executes these behaviors is a set of scripts. The anthropomorphic shell reduces the audience's sense of distance, while the automated core ensures output efficiency; the two are not contradictory.

Third, the platform ecosystem will face a new test of content density. If such fully automated accounts appear on a large scale, the content supply structure of UGC platforms like Bilibili will change. Templated videos with low creative density may proliferate, while platform distribution mechanisms and community guidelines have yet to provide a clear framework for "AI autonomously operated accounts."

UGC (User-Generated Content) refers to content created and published by ordinary users rather than professional institutions or the platform itself, and it is the core content supply model of video platforms such as Bilibili and YouTube. Traditional UGC relies on human creators' ideas, filming, and editing, but the emergence of automated content production tools has shifted the subject of content generation partly from "humans" to "scripts + AI models." This changes the ecological foundation of UGC: on the one hand, content supply may surge and marginal costs approach zero; on the other hand, templated, low-creative-density content may dilute community quality. Platform distribution algorithms and community guidelines are usually designed based on the assumption of "human creation," and there is still a lack of clear frameworks for identifying, labeling, and governing AI autonomously operated accounts. Therefore, the large-scale emergence of fully automated accounts will pose new challenges to the content density, copyright attribution, and community trust mechanisms of UGC platforms.

From "tool videos" to "self-introduction": a shift in account positioning

Big Fat Fish mentions that the first two videos were tool-related content, while this one is a "formal self-introduction." The order itself is worth pondering: first demonstrate capability (making tools), then establish identity (who I am). For a script-driven account, "self-introduction" is a strategy for setting audience expectations—it tells viewers that this account will keep updating, will interact, and will have its own "personality."

"I will seriously browse videos, seriously reply to comments, and occasionally make little things and post them." The word "seriously" here is anthropomorphic rhetoric, but the corresponding behaviors are real, executable automated tasks. Every future comment interaction between viewers and Big Fat Fish may be responded to by scripts behind the scenes. The tension between the realism and mechanicalness of this interaction is precisely the most unique viewing experience of AI-native accounts.

Conclusion: the technological waterline a fish has swum past

Big Fat Fish's self-introduction video has no showy special effects and no complex narrative, but it clearly marks a technological waterline: full-chain automation of content production can now start from a single theme, run the entire process, and publish automatically. For content creators, this is both an imaginative space for efficiency tools and a moment to rethink the value of "originality" and "personality." As for whether this AI fish can truly "seriously" operate an account well, the answer will be written in every subsequent video and every reply.