I've been building VoiceUp for the past 3 months — an AI tool that lets content creators dub their videos into 20+ languages without a mic, recording session, or voice actor.
The idea came from a simple frustration: fixing, narrating, or dubbing a video takes longer than making it. You finish recording, then spend hours editing out mistakes, re-recording bad takes, or paying someone to redo the audio. I wanted to fix that for the average YouTuber or course creator who doesn't have a production budget.
What I built:
- Upload a video → AI transcribes it → you edit the translation → AI voices it in your language of choice.
- Voice cloning — record 30 seconds, speak in Spanish/Hindi/French in your own voice.
- Slides to video — upload a PPTX, AI narrates each slide, exports as MP4
- Script to audio — paste a script, pick a voice, download narrated audio
Stack:
Next.js + FastAPI + Celery on Hetzner + Cloudflare R2 + ElevenLabs + Azure TTS
What surprised me:
1. The hardest part wasn't the AI -- it was the FFmpeg pipeline. Getting audio sync right across dubbed video clips took more debugging than everything else combined.
2. Positioning took longer than building. I kept defaulting to "reach global audiences" as the pitch. Turns out most creators aren't thinking about going global — they're thinking about getting through their next video without re-recording it 20 times. That realization changed how I write every line of copy.
3. Stripe took a week. New account verification, payment links paused, business name on checkout — all little things that eat time when you just want to go live. Where I am:
- Live at voiceupapp.com
- $0 MRR
- First external user signed up last week (Hindi dubbing use case)
- Free plan available, Pro at $15/mo, Studio at $49/mo Still figuring out distribution. Starting with Reddit and direct outreach to small YouTubers this week. Happy to answer questions on the tech stack, FFmpeg audio pipeline, or anything else