A content hub for TikTok and Douyin. Pull a creator's playlist, lift the transcript from every video, translate it, summarize or rewrite it into your own script, and generate the voice track. One chain across two markets.
Short video moves faster in Chinese than in English. A format that works on Douyin shows up on TikTok months later, so the creators worth studying were often the ones I could not skim.
Studying them properly meant watching a creator's back catalogue, understanding what each video actually said, and then turning the useful part into something of my own. Each of those is a separate tool, and the handoffs between them are where the work died: download here, transcribe there, translate somewhere else, paste into a doc, lose the thread.
Usually on Douyin, usually months before anything like it reaches an English feed.
a screenshotA separate site, an ad wall, and a file that lands in a folder under a numeric name you will not recognise tomorrow.
one at a timeAnother tool, another upload, another export, with no link back to the video it came from.
per videoPaste into a doc, lose the timing, lose the hook structure, keep the gist.
context lostThe script exists and the voice does not, so the idea waits on a session that gets scheduled next week.
the thread diesThe chain already existed as four separate products. Assembling it by hand is what made the research stop happening.
The file, for free, from any public post.
Ad walls, batch limits, and filenames that lose which creator the video came from before you have finished saving it.
RejectedAn accurate script, and accuracy is the part that actually matters here.
Every upload is manual, the output carries no link back to the post it came from, and per-minute pricing turns a whole catalogue into a decision rather than a default.
RejectedDashboards of views, engagement and follower growth, with trend charts on top.
They report how a video performed and withhold what it said. The script is the part worth studying, and the entry price sits around fifty dollars a month.
RejectedEach product owns one link of the chain and none of them owns the handoff. Keeping the source, the transcript, the translation and the final script attached to each other is the entire value.
Point it at a creator's playlist and it pulls the catalogue, lifts the transcript out of each video, and keeps the set together as a library rather than a pile of downloads.
From there the transcript is working material. Translate it, summarize it, or rewrite it into your own script in your own voice. Text-to-speech is built in, so an approved script leaves as an MP3 ready for the edit rather than as a document waiting for a recording session.
The point of keeping all of it in one place is that the chain never breaks. The idea, its source, the translation and the version you actually shipped stay attached to each other.
A creator's playlist on TikTok or Douyin becomes a tracked catalogue instead of a browser tab you lose.
Every video gives up its script, which is the part worth studying and the part no thumbnail shows you.
Cross-language by default, so a Douyin format is readable months before it reaches an English feed.
Summarize for the pattern, or rewrite into your own script. The output is yours rather than a subtitled copy.
Text-to-speech turns the approved script into an MP3, so the handoff to editing is a file rather than a booking.
Transcription runs on Gemini 2.5 Flash off the downloaded audio, which is what makes the whole-catalogue pass affordable rather than a per-video decision. AI Polish sits on top as a prompt you write yourself, so translating a Douyin script into natural English and adapting its tone happen in the same step. Everything exports to Excel or TXT, because the destination is usually a brief, a content calendar or a client report rather than the tool itself.
Every tool in this category has the same two problems, and both of them are product decisions rather than engineering ones. The first is whose account does the collecting. Shared account pools are how these tools get their users banned, so this one runs on the user's own browser session and says so on the landing page.
The second is what gets stored. The database holds the reusable work product: post metadata, source links, transcripts, polished scripts, job status and export-ready text. Source video files and generated audio stay out of it, downloads are regenerated from the source link on demand, and session values are used for collection without being written into job history. Keeping research records rather than a video library is what keeps the storage bill and the legal surface both small.
Pricing follows the same logic. A free tier at 20 videos a month is enough to study one niche, Lite at nine dollars adds AI Polish, and the top tier adds the TTS voiceover. The comparable tools in this space start around fifty dollars a month, and most of that price buys a dashboard rather than an export.
TikTok Miner answers questions about a catalogue: it embeds everything into a store you can interrogate, which is the right shape when you want to know what a creator does and why it works.
Hyper Creator is the other half. It assumes you already know what you want to make and carries one idea through translation, rewriting and voice until a file exists. Miner is for understanding, Hyper Creator is for shipping, and the same pairing shows up elsewhere in this set: AdRadar reads the market, CreativeOS writes the response.
Ideas stopped being screenshots in a folder. A format spotted on Douyin can be read, understood, adapted and voiced without leaving the tool, which is the difference between watching a market and borrowing from it.
It started as tooling for my own content and ended up packaged as a product, which is the pattern behind most of the things on this page: build it because the manual version is intolerable, then find out other people have the same problem.
ReactThe project workspace and transcript library.
TypeScriptAcross the app and the job pipeline.
ViteBuild tooling.
TailwindThe design layer.
ExpressThe job API behind the collection runs.
PlaywrightDrives the user’s own browser session. No shared account pool.
FFmpegPulls the audio track out before anything is transcribed.
Gemini 2.5 FlashTranscription off the extracted audio, plus the AI Polish rewrite and translation passes.
Neon PostgresJob status, transcripts and polished scripts. Source video files stay out of it.
Drizzle ORMSchema and queries.
XLSX exportThe actual deliverable for most users.
ResendJob-complete notifications.
StripeFree, Lite and Pro tiers with a credit meter.
VercelHosting.
Running on the user’s own session and keeping source video out of the database are the two decisions that shaped the stack. Both cost convenience and both keep the product defensible.
Nine systems shipped, compliance, reporting, intelligence, content.