The Voice Library Is Live — and Economy Delivery Halves Your Bill
A voice is an asset. It should behave like one.
Until now, every dubbing job started from zero: upload a clip, the pipeline clones whoever is talking, done. That works for one-off videos — it quietly fails for series. The narrator you cloned last month should still be available this month; the guest host of your interview format shouldn't need to re-record a sample every episode; and sometimes the voice you want doesn't exist yet and you just want to describe it.
So the library got three doors in:
- Clone from 15+ seconds of clean speech. Timbre, pitch range, delivery — carried across all nine languages.
- Design from a sentence. "A warm documentary narrator with a slight rasp." The engine renders you three takes; you audition, pick one, and freeze it as a permanent library voice.
- Assign per speaker, not per job. Multi-person videos now arrive with speakers auto-detected (S1, S2, …). Pick a voice once per person and it fans out across their lines; then override any single line you disagree with — per-line billing means a reassignment only re-renders that line.
Economy delivery: what you're actually trading
GPU capacity has a tide chart. Overnight, our render fleet is idle; by day, queues form. Economy delivery is the honest answer to that curve — the same pipeline, the same QC, no watermark, scheduled onto off-peak capacity instead of the front of the queue. The bill drops by exactly half.
| ⚡ Standard | 🌿 Economy | |
|---|---|---|
| Delivery time | minutes | within 48 hours |
| Credit cost | full | half — applies to every language and delivery option |
| Quality / QC / watermark | identical | identical |
| Line-level review, redubs, glossary | yes | yes — redubs inherit Economy |
| Who should use it | time-sensitive drops; brand-new series still tuning its voices | back-catalogues, batches, multilingual rollouts, planned series |
The math that makes it possible is boring and real: off-peak silicon costs us a fraction of on-demand capacity, so we're passing most of that saving into the price instead of the margin. If you dub a 10-minute video into Spanish and Mandarin, that's 20 credit-minutes — 10 at Economy pricing. On Creator, that's the difference between $7.20 and $3.60 per video.
Pricing 2.0, because the entry door was too high
Beta users told us the same thing twice: the craft features stick, but $29 was a lot to discover them. So the whole ladder moved down and the minutes moved with it:
- Starter — $7 / 15 credit-minutes (was $29 / 60)
- Creator — $29 / 80 credit-minutes (was $79 / 200)
- Studio — $79 / 250 credit-minutes (was $199 / 600)
- Overage $0.79 / finished minute. Credits still roll over forever; subtitles-only still costs half; failed-QC re-dos are still never billed.
Stack Economy on any plan and the effective rates get silly-fast: ≈$0.23/min on Starter, ≈$0.18 on Creator, ≈$0.16 on Studio — for a pipeline that ships per-line review, glossary enforcement and real speaker separation.
One transparency note
Economy jobs render on our own GPU facilities in Asia during off-peak hours, which means media can transit our global processing infrastructure before your US-region storage settles the master. It's the same pipeline, container-for-container identical to the one running in managed US facilities — but "where" is a fair question, and the privacy page and ToS now say it plainly. If a client contract pins your footage to one jurisdiction, use Standard; if it pins your budget, Economy is exactly why it exists.