Voxtakes Launches on Product Hunt on November 4: Per-Line AI Dubbing
Update, November 6: the launch is live and the first 48 hours are being spent on delivery capacity, not on features. Standard (cloud GPU) rendering is backfilling after a billing outage on our provider, and the economy lane on our own hardware is running in operator-started sessions. If you try Voxtakes while both lanes are warming up, tell us — that queue is exactly what we are fixing.
Why this exists
We started Voxtakes because of a single mispronounced brand name. A dub had shipped with our own product name rendered as something else — not a translation bug, a pronunciation bug. The fix was not a re-record of one line. The fix was to re-render the whole video, because that was the only unit the pipeline could bill and the only unit the tooling could re-run.
That was the moment the whole design changed. If one wrong word can cost you a full render, then review is a cost centre instead of a quality step, and every creator quietly learns to skip review. Per-line assembly and per-line billing exist to move review back into the cheap column.
We launched Voxtakes on Product Hunt on November 4, 2026 at 12:01 AM PT. If you have ever rage-quit an auto-dub, the listing is here and stays open after launch day: Voxtakes on Product Hunt. The honest version of what the launch did: it gave us a permanent product page and a maker forum, not a flood of traffic. What actually moves is answering the specific question a creator is asking right now.
What per-line dubbing means
Three things, and they are related rather than three separate features:
- Lines are the unit of assembly. A dub is a set of addressable lines, not an opaque finished file. Correct one and the correction is the new version of that line; the rest of the track stays exactly as approved.
- Lines are the unit of billing. A corrected line costs one credit-minute. You are not paying for the video you already approved. This is the single biggest difference from per-minute metering, where a four-second correction bills like a full minute.
- Lines are the unit of QC. Because each rendered line is checked against its glossary entry and its transcript, errors are attributable. A green check marks a line whose locked terms survived; an amber flag marks the line that needs a human. Nobody has to guess which sentence broke.
The reasoning behind the credit model is written up in why one bad sentence should not cost a full render, and the price comparison against metered tools is in how much AI dubbing costs in 2026 (vendor prices observed October 2026 — treat them as a dated snapshot).
What's in the box
- Brand glossary. Lock a term once with a global default, override it per language when you need to, and let the translator obey across every target. One glossary across languages instead of one dictionary per language pair — the reasoning is in one glossary, nine languages.
- Burned-in subtitle erasure. OCR finds the text, an inpainting pass repaints the background underneath, and your new captions land on clean video instead of ghost text. The ghosting failure mode and the frame-propagation fix are covered in why AI subtitle removal leaves ghosting.
- Caption presets. Pick a look at upload — clean, bold pop, word highlight, bilingual — and every dubbed language ships in that style, or export SRT/ASS and keep styling in your NLE.
- Voice library. Clone a voice from 15 seconds of clean speech or design one from a description, then assign it per speaker or per line. Speaker detection tags lines S1, S2, and so on.
The billing question
Two questions decide what a dubbing tool actually costs you: does a failed render cost me? and what does one corrected sentence cost?
Ask those two first, before comparing any two per-minute prices on any pricing page — including ours. Render-metered vendors bill the render, so a mistake made in your timeline becomes a line item, and a correction made in review becomes a second line item. Our model is the other way round: failed QC re-dos are free, and a corrected line is one credit-minute. If your pipeline produces a lot of corrections — and any reviewed pipeline does — that difference compounds long before the per-minute rates do.
Launch details
Here is the practical part.
- Launched: November 4, 2026, 12:01 AM PT (4:01 PM Beijing time, 9:01 AM in San Francisco). It is live now — the listing is at producthunt.com/products/voxtakes.
- Where: producthunt.com/products/voxtakes — comment there if you want the team to answer with what actually happens under the hood. We read every one.
- Try it: the free trial is 3 finished minutes and one watermarked video, unlocked by verifying your email. No card, no invite queue.
- If you want to hear one line before you decide: drop a video link on the homepage and the team dubs one line and emails it back — free, no signup.
One ask: if you have been burned by auto-dub, the most useful thing you can do on launch day is not just an upvote. Tell us what broke. Those threads are how the next version of this gets built.
Frequently asked questions
When does Voxtakes launch on Product Hunt?
We went live on November 4, 2026 at 12:01 AM PT — 4:01 PM Beijing time. The listing, votes and the maker forum stay open afterwards.
What does per-line billing actually mean?
Correcting a line costs one credit-minute on the line you changed, instead of re-rendering the whole finished minute. Failed QC re-dos are free, so you are never billed for our mistake.
Is the trial actually free, and do I need a card?
Verifying your email grants 3 minutes and one watermarked video. No card is required, and there is no approval queue.
See the launch on Product Hunt →Estimate your per-line cost
Related: AI dubbing cost, 2026 · HeyGen vs Voxtakes · Rask vs Voxtakes · Per-line billing · One glossary, nine languages