曾几何时,电视在人们心目中,还是个开机就能看、按两下遥控器就能换台的物种。
现在,打开电视准备看部电影,首先映入眼帘的却是一段开机广告。好不容易进了主页,扑面而来的又是满屏的海报建议、会员充值弹窗,中途按个暂停还会跳出一整张促销贴片。

最近,有一群人终于不打算再忍下去了。他们找来了一个意料之外的帮手——AI,试着把电视给「改回去」。
给电视做一场「大扫除」
上周,开发者 Gathin 在社交平台上发了一组给父母改电视的照片。按他的描述,家里的小米电视连打开 Netflix 前都要先看一段赞助广告。
于是,他把电视型号交给了编程助手 Claude Code,让它借助 Wi-Fi 接入设备,一口气停用了 25 个平台组件,再写一个极简桌面——原来杂乱的主页,一下子就变得清爽起来。
他随后把这套 prompt、桌面源码和踩坑记录开源到了 GitHub,项目命名为 CleanYourTV。
短短几天,项目迅速收获了约 1.37 万颗 Star,「天下苦智能电视久矣」了属于是。在 Reddit 的相关讨论区中,挤满了追问自家机型能否适配的使用者,不少人也陆续贴出自己改造后的极简首页。

AI「接管」电视的底层逻辑并不神秘,核心依旧是安卓生态里用了十几年的老工具:ADB(Android 调试桥)。
只要连上同一局域网、启动电视的调试权限,AI 就能顺着终端接口排查正在运行的几百个应用组件。哪些是开机广告,哪些是偷偷上报收视数据的遥测程序,AI 逐个筛查并下达命令将其「冻结」,最后再换上一个纯净的开源桌面。
甚至原本深藏在多层二级菜单里的 HDMI 与外接输入源,也能直接做成首页的常驻入口。

另一位 YouTuber Ansub Khan 也记录了自己的改造过程,并将项目开源为 TV Detox。在他原本的那台电视上,开机默认会自动播放嘈杂的轮播内容。
改造之后,新首页只剩下时钟、常用视频应用和输入源。

视频里,Ansub 实测电视可用内存从约 370 MB 飙升到了 600—650 MB;一位照着 prompt 操作的 TCL Beyond 使用者也表示,虽然没有精确数据,但摆脱了沉重的后台驻留后,整台电视的滑动与响应显著变轻快了。
翻看 Reddit 社区里的讨论,我注意到一个很有意思的细节:人们心目中想要的「理想电视」,似乎根本不是同一回事,彼此甚至完全无法达成一致。
有人仍旧习惯原厂界面,只想剔除广告和多余建议,再把遥控器的快捷键重映射给常用应用;
另一位索尼 KD-55A8H 使用者则更追求极简——他把时钟改成了 24 小时制,删掉早晚问候、日期和所有装饰线,连平台字体都换成了工整冷峻的 IBM Plex Sans。

事实上,人们的需要本来就是千人千面的。但在追求标准化的商业逻辑里,使用者往往都只能在厂商给定的框架内自定义。
戏剧性的是,AI 的介入反而意外地给这种极度细分、千奇百怪的个性化诉求,给出了一条普通人也能摸索走通的定制路径。
硬件不赚钱,显示屏就成了广告位
这些年电视越来越便宜,除了面板成本下降与生产力提高,也有厂商「赔本赚吆喝」的成分在。但厂商毕竟不是做慈善的,硬件负责低价「交个朋友」,钱自然就得从别的地方赚回来。
至于电视卖出去之后还能靠什么赚钱,笔者可以从海外流媒体巨头 Roku 的财报中略窥一二。
2025 年,它的设备硬件业务全年毛亏损了 8202 万美元;依靠数字广告、订阅与内容分发构成的平台业务,毛利润却高达 21.56 亿美元。

TCL 电子也披露,2025 年互联网业务收入为 31.1 亿港元,毛利率达到 56.4%;同期,大尺寸表明业务的毛利率为 16.8%。很显然,围绕电视开展的互联网运营,有着比硬件更厚的毛利空间。
在国内,针对智能电视的整改确实有了一定的成效。依据广电总局披露的数据,主流品牌已有 93% 的机型兼容将开机广告一键关闭或永久关闭。
但这仍旧治标不治本。开机倒计时广告被限制,并阻挡不了厂商变现的需要,商业算盘只会打得更加隐蔽。

就拿国内的一些流媒体平台来说,片头动辄 90 秒的贴片广告、暂停时的全屏促销海报,以及手机端年卡不可用的「大屏专享 VIP」弹窗,仍旧像牛皮癣一样原封不动地贴在那里。
CleanYourTV 这类工具能清理的,仅仅是电视平台自身的预装包与官方桌面。它能帮你把客厅大门打扫干净,却管不到各个 App 屋内的规矩。
笔者买下的硬件,为什么不能自己做主?
「Vibe-Hacking」这股风潮,并不只局限在「电视」。
在相机圈,一群摄影师也在想方设法把厂商藏起来的能力找回来。
在索尼老机身上,开源项目 Recipe Lab 借用老旧的 PlayMemories Camera Apps 入口,让 A6000、A6300 等机型可以加载自定义色彩配方,把色彩矩阵与白平衡参数写入机身底层。
另一项针对 A7 IV 固件的研究,更是在 6.02 版本固件深处直接翻出了被官方屏蔽的快门角度与 LUT 达成。虽然当下仅能调出图标、拨轮调节仍旧受限,但代码的存在本身就证实了硬件能力早已有之。

尼康社区的开源项目 Wonder Lantern 则把个性化做到了更深处。有开发者不仅为 Z30 扩展出了 900 秒超长曝光、直方图同屏以及自定义快门音,国内逆向社区甚至直接解包并重构了尼康固件,借助私钥重新打包签名,成功在相机菜单里加入了新的设置选项。

理光社区的 GR Mod 则把摄影师自己的 .cube LUT 色彩预设,拟合成相机能识别的机内风格槽位,甚至加入了 65:24 宽幅比例,让这台街拍机直接具备了类似胶片风格与 XPAN 构图视野。

在富士圈子里,开发者 justice-jake 正在试图打破官方的固件护城河——把新款机型独占的胶片模拟、快照对焦和配方编辑器,反向移植到快被遗忘的老机身 X70 上。
老款处理器的通道曲线被硬件锁死,他就借助 AI 辅助 USB 调试与固件仿真,在底层代码里一点点抠出近似的算法达成。

上下滑动查看更多内容
更具戏剧性的探索浮现在哈苏身上。开源项目 Hasselblad X Unleashed 直接瞄准了部分中画幅机身的底层平台——这台相机内部跑的是深度定制的 Android,这扇隐秘的「后门」让极客们兴奋不已。
他们不仅顺藤摸瓜解锁了无线麦克风、外接监看和像素位移,社区甚至有人直接在价值数万元的哈苏相机上成功跑起了经典游戏《毁灭战士》。

与在电视上停用几个安卓软件相比,去啃相机底层的固件显然是完全不同量级的冒险。
但这两群看似风马牛不相及的人,折腾的逻辑异曲同工:笔者明明花全款买下了这台机器,它的能力到底该怎么用,凭什么只能由厂商的商业算盘说了算?
过去,想要突破这种设定,需要极客在汇编代码和反编译工具里硬啃数月甚至数年。
现在,AI 第一次把原本专属于逆向工程师的代码审计和排查能力,交到了更多普通人手里。那些原本被封装在黑盒深处,被刀法阉割的能力、塞满广告的角落,第一次变得透明可见。
从客厅里的显示屏大扫除,到摄影圈里的固件解锁,一场自下而上的工艺反叛,正在一次次看似微小的折腾里显出轮廓。
商业公司的盈利算盘不会停止,数字围栏也不会凭空消失。但当工艺的天平悄然发生倾斜,那些曾经被视作理所当然的出厂设定,终于被重新打上了问号。
欢迎来到「大逆向时代」。
本文来源:https://www.ifanr.com/1685473
若有版权问题,请来信告知:13951431913#139.com 或在右方提交评论。
Once upon a time, television was still that simple creature you just turned on to watch and switched channels with a couple of clicks.
You turn on the TV to watch a movie, only to be greeted immediately by an boot-up advertisement. Just when you finally manage to reach the home screen, you are bombarded with wall-to-wall poster recommendations and pop-ups begging you to sign up for a membership. Even if you pause mid-stream, a large promotional banner will still suddenly appear across the screen.
Recently, a group of people decided they could no longer put up with it. They turned to an unexpected helper—AI—to try and "change it back.
Last week, developer Gathin posted a series of photos on social media about modifying the TV for his parents. He explained that their home Xiaomi TV forces you to watch a sponsored advertisement before you can even open Netflix.
So, he handed the TV specifications to the coding assistant Claude Code, which accessed the device via Wi-Fi, disabled 25 platform components in one go, and wrote a minimalist desktop—instantly transforming the cluttered homepage into something clean and fresh.
He then open-sourced these prompts, desktop source code, and troubleshooting logs on GitHub, naming the project CleanYourTV.
In just a few days, the project quickly garnered around 13,700 stars. It really validates the sentiment that people have been suffering from smart TVs for far too long. Related subreddits are flooded with users asking if their specific models are compatible, and many have been sharing photos of their modified, minimalist homepages.
The underlying logic behind AI taking over televisions isn't mysterious; it still relies on an old tool that has been used in the Android ecosystem for over a decade: ADB (Android Debug Bridge).
As long as it is connected to the same local network and debugging permissions on the TV are enabled, the AI can audit the hundreds of running application components via the terminal interface. It identifies boot advertisements and telemetry programs that secretly report viewing data, screening them one by one and commanding them to be 'frozen'. Finally, it swaps in a clean, open-source desktop.
Even HDMI and external input sources, previously buried deep in multi-level submenus, can be set as permanent, quick-access options on the home page.
Another YouTuber, Ansub Khan, also documented his modification process and open-sourced the project as TV Detox. On his original TV, turning it on would automatically play noisy carousel content by default.
After the redesign, the new homepage consists only of a clock, commonly used video apps, and input sources.
In the video, Ansub's hands-on test showed that the TV's available memory surged from roughly 370 MB to 600–650 MB. A TCL Beyond user who followed the prompt also reported that while they didn't have precise data, after clearing out heavy background processes, the entire TV's scrolling and responsiveness became noticeably snappier.
Browsing through Reddit discussions, I noticed something quite interesting: people's visions of the "ideal TV" seem to be completely different from one another, and they can't even reach a consensus on what it should be.
Some people still prefer the stock interface and just want to remove ads and unnecessary recommendations, then remap the remote's shortcut buttons to their frequently used apps.
Another Sony KD-55A8H user, however, is even more minimalist—he changed the clock to a 24-hour format, removed the morning and evening greetings, the date, and all decorative lines, and even swapped the platform's font for the clean, stark IBM Plex Sans.
In reality, individual needs are inherently diverse. However, driven by the logic of standardization, users are often restricted to customizing within the confines set by manufacturers.
Paradoxically, AI's intervention has unexpectedly provided a viable customization path for these hyper-specific and diverse demands—one that ordinary people can actually navigate and follow.
TVs have become increasingly affordable in recent years. While this is due in part to lower panel costs and increased productivity, it also stems from manufacturers' willingness to sell at a loss just to generate hype. However, manufacturers are not running charities; they use the hardware to offer low prices and "make friends," but naturally, they have to make their money back elsewhere.
As for how to make money after selling TVs, one can gain some insight from the financial reports of the overseas streaming giant Roku.
In 2025, its hardware business reported a full-year gross loss of 82.02 million US dollars; however, the platform business—consisting of digital advertising, subscriptions, and content distribution—generated a gross profit as high as 2.156 billion US dollars.
TCL Electronics also disclosed that internet service revenue for 2025 reached HK$3.11 billion, recording a gross margin of 56.4%. In comparison, the gross margin for the large-size display business was 16.8% for the same period. It is evident that internet operations centered around televisions offer significantly higher gross margins than hardware.
Domestically, efforts to regulate smart TVs have indeed shown results. According to data released by the State Administration of Radio and Television, 93% of models from major brands now allow users to skip startup ads with a single click or disable them permanently.
However, this is merely a superficial fix. Restricting boot-up countdown ads will not halt manufacturers' need to monetize; they will only grow more covert in their business calculations.
Take some domestic streaming platforms, for example: 90-second pre-roll ads, full-screen promotional posters during pauses, and pop-ups advertising "Big Screen Exclusive VIP" tiers that aren't valid for mobile annual passes—these annoyances remain stuck there like a stubborn rash.
Tools like CleanYourTV can only remove pre-installed packages and the official home screens provided by the TV platform itself. While they help tidy up the front door, they cannot control the rules or inner workings within individual apps.
The "Vibe-Hacking" trend is not limited to just "television.
In the camera world, a group of photographers is also trying to recover the capabilities that manufacturers have hidden.
Through the legacy PlayMemories Camera Apps interface on older Sony camera bodies, the open-source project Recipe Lab enables models like the A6000 and A6300 to load custom color recipes by embedding color matrix and white balance parameters directly into the camera's underlying firmware.
Another analysis of the A7 IV firmware has unearthed officially disabled shutter angle and LUT functions deep within version 6.02. While currently only the icons can be accessed and dial adjustments remain restricted, the mere existence of this code confirms that the hardware capability has long been present.
The open-source project Wonder Lantern in the Nikon community takes personalization to an even deeper level. Not only have developers extended the Z30 with 900-second ultra-long exposures, on-screen histograms, and custom shutter sounds, but the domestic reverse engineering community has even directly unpacked and restructured Nikon's firmware, repackaged and signed it using private keys, and successfully added new settings options to the camera's menu.
The GR Mod for the Ricoh community allows photographers to use their own .cube LUT color presets, emulating them as in-camera style slots the camera can recognize, and even adds a 65:24 panoramic aspect ratio, giving this street-shooting camera a film-like style and the wide field of view of an XPAN.
In the Fujifilm community, developer justice-jake is trying to break the official firmware moat by backporting features exclusive to newer models—Film Simulations, Snap Focus, and the Recipe Editor—to the nearly forgotten X70.
The channel curves of the older processors were locked down by hardware, so he leveraged AI to assist with USB debugging and firmware emulation, working out, bit by bit, approximate algorithms from the low-level code to recreate the intended behavior.
Scroll down for more content
A more dramatic development has surfaced with Hasselblad. The open-source project Hasselblad X Unleashed directly targets the underlying platform of certain medium format camera bodies—internally, these cameras run a deeply customized version of Android. This hidden "backdoor" has excited geeks everywhere.
They not only figured out how to enable wireless microphones, external monitoring, and pixel shift, but someone in the community even managed to run the classic game *Doom* on a Hasselblad camera worth tens of thousands of yuan.
Compared to disabling a few Android apps on a TV, tackling the low-level camera firmware is obviously a completely different level of risk.
Yet, for these two seemingly disparate groups, the underlying logic behind their struggles is identical: I paid in full for this machine, so why should its capabilities be dictated solely by the manufacturer's commercial calculations?
In the past, to break through such limitations, a geek had to slog through assembly code and decompilation tools for months or even years.
For the first time, AI is putting the code auditing and troubleshooting capabilities once reserved for reverse engineers into the hands of ordinary people. What was previously hidden deep within black boxes—features crippled by artificial limitations, and corners cluttered with ads—is now becoming transparent and visible.
From overhauling living room displays to unlocking firmware in photography circles, a bottom-up rebellion in craftsmanship is taking shape through repeated, seemingly minor tinkering.
Corporate profit calculations will not cease, nor will digital fences vanish into thin air. But as the balance of craftsmanship quietly tips, those factory settings once taken for granted have finally come under renewed scrutiny.
Welcome to the Great Reversal Era.
Source: https://www.ifanr.com/1685473 If you have any copyright concerns, please contact us at 13951431913#139.com or submit a comment on the right.