Perso AI

Perso caps each video by plan before credits matter. A single upload can run 3 minutes on Starter, 15 on Creator, 30 on Pro and 60 on Enterprise. A 20-minute lecture therefore needs Pro even if the monthly balance on Creator would cover it, or it must be split into shorter parts before upload. Free is separate: one video of up to a minute, generated once per new account.

Monthly dubbed minutes then decide how many videos fit: 7 on Starter, 40 on Creator and 180 on Pro. Languages widen with the plan, from 12 supported dubbing languages on Free to 74 on Starter and Creator and 111 on Pro. Editorial assessment: Creator suits someone dubbing a handful of 5 to 15 minute videos into one or two languages each month, while Starter is mainly a trial for short clips.

The comparison table lists lip dubbing as its own monthly allowance: 2 minutes on Starter, 13 on Creator and 60 on Pro. That is a fraction of the dubbing minutes on the same plan. A creator who wants every translated minute lip-synced should compare the runtime with this smaller number, not with the headline credit total.

Pro adds the Nightingale model at 6 credits per second, or 360 credits per minute, next to the Oriole and Wren models on lower plans. The lip-sync page describes frame-level processing that copes with mouths partly covered by hands or microphones. Those are provider claims, and a short clip of your own footage is the way to check them.

Paid plans list more than the dubbed MP4. The comparison table includes source and translated audio as WAV, per-speaker source audio as a TAR archive, background music as WAV, source and translated subtitles as SRT, and scripts as XLSX. An editor can rebuild the mix or subtitle track in another application instead of accepting the rendered video as final.

Resolution rises with the plan: 720p on Free, 1080p on Starter and Creator, and 4K on Pro. Storage changes too: Free projects are kept for 30 days, paid plans without a stated limit. Download what you need promptly on Free, and keep the WAV and SRT files with the delivered video in case a client asks for corrections later.

Voice cloning and lip sync reproduce a real person, including voice and mouth movement. The pricing FAQ makes you responsible for holding rights to the source material, including voice and image rights, and the voice-cloning page states that models are created only with user consent. For anyone other than yourself, get written permission that names dubbing, cloning and the target languages.

The terms treat facial and voice data as biometric information: encrypted, stored separately, destroyed after its purpose and deletable on request. They also let ESTsoft use member data for AI learning; refusing is possible but may restrict some services. The trust page adds that Perso belongs to the Content Authenticity Initiative and supports metadata that labels AI-dubbed content.

Subscription credits are valid for one month from issue, separately purchased credits for one year, and the earliest-expiring credits are spent first. When credits are exhausted, new dubbing stops, but existing projects can still be regenerated, viewed and downloaded. Creator and Pro users can keep generating in a slower reduced-speed mode.

Cancellation takes effect at the next billing date, with access until then. The terms allow withdrawal within seven days of payment, limited for credits already used, and otherwise make fees non-refundable. They also prohibit editing downloaded content into derivative works and saving it other than through the download feature, so confirm with ESTsoft whether post-editing a delivered file fits your contract.

1. Upload the clip and note whether the glossary is offered on Free.
2. Review the transcript and translation, and correct one line deliberately.
3. Generate, then watch the pause, the brand name and the final sentence.
4. Check whether lip sync was applied or shown as unavailable on Free.
5. Download the MP4 and note resolution, watermark and remaining balance.

Accept a dub whose timing stays within the clip, pronounces the brand name correctly and keeps the voice recognisable. Reject cut-off speech, a mistranslated corrected line or unusable mouth movement. No result is recorded until the task is run.

For presenter videos built from an avatar rather than your own footage, see HeyGen or Synthesia. ElevenLabs centres on voice work, and Dzine on image-driven lip sync. Compare the same clip before comparing price.