yeah the caption lock at export is rough. if you only need generate → tweak → export (or burn an SRT so it sticks on tiktok/reels), this browser path works for me: https://t.co/SwsSc17YfS — burn: https://t.co/G3dCY9By1b
kalau cuma butuh auto caption tanpa CapCut Pro, bisa di browser: generate → edit → export. free di https://t.co/SwsSc17YfS — udah punya SRT tinggal burn biar nempel di tiap app https://t.co/G3dCY9By1b
ges ada yang bisa tambahin auto captions ke video aku ga? drop fee dong plz cm 2 menit videonya.. pgn auto caption tp capcut tdk pro :) yg bisa lgs kelar skrg #zonauangㅤ#zonaba
shipped a new page on subtitlegenerator today: burn subtitles right into your video so they stay put on tiktok, reels, wherever. codex built it, claude reviewed it, i said yes at the end. 430 unit tests + 91 browser checks before it went live https://t.co/G3dCY9By1b
most replies here are designers for hire — if you want DIY instead, i’ve been building ThumbMuse for “i know the idea, just need 2–3 thumbnail directions.” drop a title/YT link and it gives strategies + options. curious what “great” means for you — loud face+text CTR, or cleaner?
is it the glossy face / warped text look, or more that it usually signals low-effort video? asking cause i keep seeing the same “ai slop thumbnail” pattern and want to know what actually triggers the skip
as a non-coder who lets codex do basically all my building, the AGENTS.md-in-Downloads thing spooked me a bit
a planted file got codex to install malware 90%+ of the time in tests. new rule for me: agents only start inside a real project folder. never ~/Downloads
spent my morning learning xiaomi's share sheet has no "save to device" option. only apps. so android users could export their video on subtitlegenerator… and then not keep it lol
fix: real "save to phone" + "share to an app" buttons. tested on an actual phone this time
wait yeah mine vanished too 😭 if you need them back for now, burning captions into the video before you upload is the safest bet since they stay put on every app. subtitlegenerator does that + translation in the browser if you wanna try
ugh yeah capcut paywalling auto captions is rough. if you just need captions on a clip, subtitlegenerator does it free in the browser — free export has a small watermark tho, no pressure to upgrade. try it if you want
same. codex is great inside a fence, but left alone it loves over-engineering bugfixes into mini frameworks
what works for me now: claude orchestrates / assigns / reviews, codex just writes inside the rails
@woshiyanali Same, and I think that's the right call. My rough split: Claude Code for big messy refactors, Codex for quick bug fixes, and Grok Bot running the boring ops around the builds. Pick the tool per mess, not per year.
everyone's doing the "pick one forever" thing — claude code, codex, or cursor
as a non-coder shipping a bunch of tiny ai products… i refuse. i bounce between them depending on the mess
also cursor computer use just dropped so agents can click around now. wild week
timedsubs used to drop you somewhere random after a pay-per-use checkout — mid-edit script + audio project? gone. cold restart energy
fixed it: you land right back where you left off
https://t.co/ypd2aOOya6
fun fact: the rumpelstiltskin meme isn't from a 1987 movie, it's AI. so naturally i made an AI tool to put your face in it 🤷♀️ https://t.co/cvqgkknVGJ
I added a @WHOOP integration to TrustMRR.
It shows this data on your /founder page:
😴 Sleep & consistency
👌 Recovery, HRV, RHR
🏋️♀️ Strain & kcal burnt
The end goal is to really see if a good sleep routine and regular exercise increase founders' revenue.
The app is still in review, so there are only 8 spots left for now.
First come, first served!
Speaker labels aren't shipped, and one comment doesn't set the roadmap.
A clean two-speaker test found 4/4 turns at 0.94–1.00 confidence. Next: overlap, noise, correction UX, translation, and export.
Full recap:
https://t.co/Ttys2xsBS1
When do subtitles feel finished?
Gal Dayan's Product Hunt comment opened the next loop: speaker labeling when two people overlap.
The transcript can read correctly, but assigning each line to the right speaker remains manual work in interview and podcast clips.