Siber Sörfçü opened an Instagram account for Elara Whitmore, an AI character created from scratch, and ran a seven-day growth experiment by publishing content regularly. The Higgsfield-sponsored video follows character design, a weak start, a separate production experiment delegated to Claude, the shift from dance videos to short films, and developments after the first week. The initial question was not simply whether the character could attract views, but whether she could develop an identity capable of attracting commercial interest.
The channel’s account is not an uninterrupted success story. It includes low view counts, inconsistent generations, an older video that unexpectedly took off, and a brief account shutdown. The seven-day result and later updates also represent different stages: approximately 19,000 followers in the first week, 13 million total views and 32,000 followers in the eight-day summary, and 40,000 followers in the final Rapunzel update.
How Elara’s appearance was created
The first step was defining the character’s appearance in Higgsfield. The channel compares the template-based interface, which offers choices such as hair, facial features and skin tone, to the character creator in a role-playing game. Although generating a character directly from a prompt was also possible, these preset options provided the initial inspiration.
From several variations, the channel selected a tall female character with a particularly long neck. The intention was to create a figure whose appearance differed from the familiar human images in a social feed. A new Instagram account was then opened under the name Elara Whitmore; because the username was already taken, “official” was added. Her physical design was ready, but the kind of content she would produce and the reason people would follow her had not yet been established. Subsequent changes would affect both the videos and this original appearance.
Preparing the new account and researching initial content
Drawing on earlier experiences, the channel did not publish a video immediately after opening the account. It followed similar AI accounts, browsed Instagram and liked various videos. After a while, the account’s feed consisted largely of AI influencer content. This was the preparation method Siber Sörfçü chose for the experiment.
To find initial video ideas, the channel asked Claude to research popular content and selected examples that seemed suitable for Elara. It describes operating AI tools and designing influencer content that people will watch on Instagram as different skills. With limited time, the initial decision was to adapt videos that had already attracted interest rather than develop an entirely new format. Research, reference selection and production were therefore connected from the beginning: the first objective was to see how an established example would perform with the channel’s own character.
Replacing the person in a reference video with Elara
For the first generations, the selected video was uploaded to Higgsfield alongside an image of Elara as the character reference. The platform was asked to replace the person in the video with her. Although the channel found the result, produced within a few minutes, technically impressive, it also questioned why anyone would want to watch that particular piece of content.
After publishing the first post, the channel continued looking for ideas. For the second video, it selected footage of a woman dancing and again adapted it using both the video and character reference. The initial posting plan was two or three videos per day, with at least five hours between posts. The completed videos were therefore not uploaded consecutively. At this stage, the content revolved around dance footage remade through character replacement; the short-film storytelling that developed later was not yet present.
Early results began with 13 views
The first video received 13 views in an hour. The next counts the channel recorded were 36 and then 52. After five hours, the second post went live; the account had reached eight followers. The second video climbed to 126 views. This small increase prompted a third dance video, but the new generation did not preserve Elara’s distinctive proportions.
Following the reference movement, the model produced the long-necked figure with more conventional proportions. When a second attempt yielded a similar result, the video was published in that form the following day. It performed better than the earlier posts, quickly reaching 600 views, while the follower count rose to 13. These were still not the growth figures the channel wanted. The process had also exposed a difference between designing a striking character and keeping that character’s movement and appearance consistent across reference videos.
Making the long neck the focus of the content
The next idea was to turn Elara herself into the attraction rather than simply show her performing an action. The channel thought her long neck could make someone pause for a few seconds. It found a video that looked as though it had been filmed from a distance, showing a woman walking along a street while people around her watched, and adapted it for Elara.
Prepared as the second video of day two, the post was sitting at approximately 180 views when checked. Another video stopping at the same level led the channel to look for a pattern. In the video, this is interpreted as Instagram initially testing content with a small audience. That interpretation is Siber Sörfçü’s reading of its own results. The point that drove the next stage was that finding an explanation for low numbers did not, by itself, provide a new content idea.
A second experiment: handing ideas and production decisions to Claude
The channel then tested whether Claude could achieve a better result than its own chosen ideas. It connected Higgsfield’s MCP integration, gave Claude browser access to Instagram and requested a 15-second video suited to Elara. The intention was for the model to make the decisions, from the concept through to the title. The channel specifically told it not to ask questions.
Claude found the character, researched trends and said it would produce several versions before selecting one. It assessed the generated videos with an analysis tool and made changes, such as adding text, when it considered a result weak. The channel says the process took approximately an hour and involved numerous generations before arriving at one result. A less-directed AI workflow was thus tested within the larger influencer experiment. However, the selected concept was markedly different from the videos the channel would have prepared itself.
683 credits spent without the hoped-for breakthrough
This Claude-led sub-experiment consumed 683 credits in total. Siber Sörfçü said that, according to its calculation at the time, this amounted to approximately $30. For comparison, it put the average cost of its own directed productions at around $3 per video. While explaining the difference, the channel stressed that it had itself chosen to delegate all decisions and prohibit follow-up questions.
The resulting video was published as the final post of day two. A check that night showed that, although it appeared to perform slightly better than the previous average, it had not passed approximately 200 views. The channel presents this as a personal experience in which regular production, or asking an agent to perform more operations, did not automatically reveal the desired creative direction. Rather than continue unchanged, it returned to research, continued discussing ideas with Claude and watched tutorials from other content creators.
A new concept found in the profile description
A recurring suggestion in that research was to find a specific niche or concept. The starting point for the new idea was a profile description previously written by AI: the character would dance in places where she should not be dancing. The channel realised it could use AI to create unusual settings rather than imitate ordinary people in everyday locations.
It prepared two videos featuring unusual locations and movements specific to Elara. Her long neck would now be part of the choreography, rather than merely a visual feature. The videos were posted five hours apart, but the results remained low; the channel says they attracted fewer views than the adaptations. Toward the evening, it switched to a completely different concept. This new post received thousands of views. By the next morning, an earlier dance video had also gained momentum and the account had reached 72 followers. To explore whether the new concept had caused that movement or the older video had simply begun circulating later, it published another post in the new direction.
A short-film identity takes shape around day four
The new direction was filmmaking. Instead of only dancing, Elara starred in vertical short films lasting between 30 seconds and a minute. Each contained a self-contained story, an intriguing or surprising event, or sometimes a wholly absurd situation. The channel describes this approach as aligning with the work it enjoyed, from developing a story and designing scenes to editing them together.
As the new posts gained traction, the character’s appearance also changed. Her neck was shortened to bring her closer to normal human proportions. The video says that likes, comments and follows accelerated on day four; the subsequent account describes reaching 10,000 followers in a short time. Elara had started as a figure that could be inserted into any video. Now the intention was to give her an identity with continuity across stories. The distinction the channel highlights is between operating a tool and giving a character a personality and a reason to be watched.
A brief account shutdown and the rise of an older dance video
Just as momentum was building, the account briefly shut down. The channel says no reason was provided and that the account became active again immediately after it entered the information Instagram requested. Production continued afterward. While the short films attracted 30,000 or 50,000 views, an older dance video unexpectedly accelerated.
In that video, Elara danced in front of an Egyptian king. The channel describes seeing more comments and likes each time it refreshed the page, including humorous reactions to the character’s long neck. The post was shared to Stories, sent to friends and reposted unchanged by another account. Siber Sörfçü also says some people wrote to Elara as though she were real, despite descriptions identifying the account as an AI character. The older video’s continuing popularity prompted another reconsideration of the account’s direction while work on new short films continued.
Combining the attention from dancing with a filmmaking goal
For a while, the channel responded to that interest by producing more dance videos. It studied popular examples, created different versions and found that these also attracted views. Noting that a substantial share of the comments were in Arabic, it recognised that it was reaching a different audience from the one originally envisaged. During this period, the posting frequency increased to four videos a day.
However, the channel did not want to make only dance content. Short films offered a broader space for different characters and worlds. It therefore planned a gradual transition rather than abandoning dancing at once: some posts were dances and others were vertical films. Elara’s profile described her as an actress and belly dancer. Over time, she became a character experiencing events suited to her identity rather than a figure placed randomly into any story. The goal was to combine the channel’s enjoyment of filmmaking with audience interest and ultimately build an account that could become sustainable.
The first-week result: approximately 19,000 followers
As the seven-day experiment approached its end, the balance shifted. Dance videos began attracting less interest than before, while the short films gained more views. The channel reported 19,000 followers on the final day, rounding that to approximately 20,000 in the same account. It said most short films received between 50,000 and 100,000 views, with some exceeding that range.
Nevertheless, the Egyptian dance’s more than three million views remained an important point of comparison. Producing similar dances had not repeated its success. Siber Sörfçü wanted the films receiving its main creative effort to attract attention on that scale too. Although the first seven days did not fully deliver that objective, the account had moved far beyond its initially tiny numbers. The video then follows the developments after that first week.
Day eight: a short film jumps to two million views
The day after the experiment ended, one of the most recent short films unexpectedly gained momentum. The channel describes this as a turning point that helped it understand differences between earlier posts. It then made a single film on day eight to test what it had learned, publishing it after generation, editing and assembly were complete.
According to the channel, this film received two million views in half a day. Siber Sörfçü describes its rise as ten times faster than the Egyptian dance video, bringing hundreds of comments and shares and thousands of new followers. Siber Sörfçü links its result to creating something viewers could discuss, rather than simply offering an image to watch. A later section of the video says another film also passed two million views. The figures provided for the eight-day total are 13 million views and 32,000 followers. A single film’s views and the account’s total views therefore remain separate measures.
Different reasons to watch, comment and follow
The channel’s interpretation of the experiment goes beyond asking which video received more views. It says it began thinking first about why someone would watch a piece of content and then why they would comment on or share it. In the Egyptian dance, the long neck attracted attention, the rhythm helped sustain viewing, and the imagery gave people material for comments. The short films attempted to create that interest through the channel’s preferred form of storytelling.
Siber Sörfçü also explicitly states that the films were not all entirely original ideas from the outset. It adapted content it knew had attracted interest to suit Elara and the account’s concept, then remade it. In the channel’s interpretation, however, copying alone did not establish a lasting direction. Getting someone to watch a post and getting them to follow an account were different questions. Giving Elara a personality and connecting each post to that identity became the distinction emphasised later in the experiment.
135 videos, more than 200 images and $390
The twenty-something published posts did not represent everything produced. The channel disclosed that it generated 135 videos and more than 200 images in total. These included alternatives made while choosing the character, adjusting her appearance, testing scenes and finding the account’s concept. It calculated the total cost of the credits used at $390.
This amount is the production cost disclosed for the completed experiment. The channel thinks it could consume fewer credits if it repeated the work with what it had learned. Higgsfield’s sponsorship is also disclosed from the beginning. Siber Sörfçü says it chose to conduct a genuine growth experiment with uncertain results rather than merely demonstrate the platform’s features. Failed generations and changes of direction are consequently included in the reported cost.
The Rapunzel update: eight million views and messages from three companies
An update added to the end of the video describes a Rapunzel short film based on the channel’s own idea and suited to Elara’s identity. The channel says that by the following morning, the film had reached eight million views and the account had reached 40,000 followers. This is a separate stage after the eight-day figures of 13 million total views and 32,000 followers. In its closing assessment, the channel describes the transformation as a nine-day process.
The same update reports sponsorship messages from three genuine companies. Two were AI companies; the channel says they wanted to support credits and content production, with one proposing a series. One of them was Higgsfield, already the sponsor of the video. The channel says the company contacted Elara’s account without knowing it was connected to the creator behind the experiment.
The third message came from a beauty brand. Siber Sörfçü relates this to the type of brand it had wanted to attract while changing Elara’s appearance and presentation. In closing, it says it transformed the account into a “money-making” one in nine days. The reported outcome is a growing audience and incoming commercial interest. A separate follow-up tracking Elara’s actual earnings remains a possible future project in the account. The channel proposes making that follow-up if the current video reaches 10,000 likes.
