Visual Translate

Positioning: the layer every other video translator ignores

Most video translation products do two things: swap the spoken audio and burn translated subtitles. Both sit on top of the picture. What they leave alone is the text baked into the picture itself — slide titles, product callouts, safety warnings, lower thirds, chart labels. Visual Translate, a feature of the Vozo video platform, exists to attack exactly that layer: it detects on-screen text in a video, erases it, translates it, and rebuilds it in the target language while preserving layout, style and animation, with no access to the original project files required (sources: vozo.ai/visual-translate, Vozo help centre). In my testing of localisation workflows, this is the single most common source of half-localised deliverables: a training video where the narration is in German and the equipment labels are still in English. It is a narrow job, but it is a job that almost nobody else does as a standalone product — Vozo markets the API as available only from them (source: vozo.ai/api).

What it actually does

The pipeline is three steps, and the product is honest about the order. Detect: it finds the visual text viewers actually read — slides, labels, titles, annotations. Translate: with context, optionally steered by a custom translation prompt and constrained by glossaries so terminology stays consistent across a project. Rebuild and proofread: the original text is erased and the translated text is composited back in its place, then you review side by side — original and translated video playing together — and edit before export (sources: vozo.ai/visual-translate, Vozo docs 'Get Started'). The editor is real, not a token gesture: any text element can have its content retranslated, its font, size and colour adjusted, and its in/out animation and timing (start, duration, layer order) moved. That animation-and-timing control is what separates this from a subtitle overlay; the rebuilt text animates like the original did.

The language coverage is broad: 58 source and 165 target languages for Visual Translate specifically (the audio dubbing feature supports 111 source languages), and the target list runs from regional Arabic and Chinese variants to Welsh and Zulu (source: vozo.ai/visual-translate FAQ, vozo.ai/pricing). Resolution support is up to 4K input with 1080p output, and the hard cap is 20 minutes of video per file on every tier, including Enterprise (source: vozo.ai/pricing feature table). Cost is metered in AI points at a flat 10 points per minute of uploaded video (source: Vozo docs, 'AI Points Rules'). The plans those points live on, at time of writing: Free with 20 trial points (roughly 2 visual translate minutes) and a 3-project limit; Creator at $29/month; Studio at $99/month with 600 points (about 60 visual translate minutes), 3 seats and bulk upload; Studio XL and XXL above that with 1,500 and 4,000 points; Enterprise by quote with human review as an add-on (source: vozo.ai/pricing). Points from monthly subscriptions carry over one month (valid two months from issue), and point packs bought separately last a year; points only burn while a subscription is active (source: Vozo docs).

There is a workflow continuation worth knowing about: a finished Visual Translate project can proceed to subtitles, dubbing and lip sync within Vozo, and the docs say a unified single-workflow integration of all four is in progress (source: Vozo docs 'Get Started'). API access is included from the Creator plan up, sharing one point balance with the web app, with a default 10 concurrent jobs per API type and automatic point refunds on failed jobs (source: Vozo API docs).

My testing notes

Two concrete observations, plus one on the cost model.

First, the detection and rebuild on clean slide content is the product at its best. I ran a 4-minute product walkthrough with heavy text overlays — feature callouts, a pricing slide, animated lower thirds — and the output rebuilt all of it in Spanish with the source styling intact: same font weight and colour on the callouts, and the lower-third animations played with the translated strings. The side-by-side preview earned its keep here: one callout mistranslated a product term that the glossary should have caught (the glossary entry existed but was applied at project level and the segment came from a nested overlay), and the editor's retranslate-one-element flow fixed it in under a minute without re-rendering the video. On a second pass with a training video containing safety warnings, the text was replaced accurately but the rebuild occasionally produced slightly tighter line-wrapping on longer German strings — adjust the font size in the style panel and it is fine, though it is manual work per element.

Second, the documented boundaries matched what I saw, which is rarer than it should be. The product explicitly excludes UI elements (software interfaces, buttons, menus) and real-world text (packaging, signage, text on physical objects) (source: vozo.ai/visual-translate FAQ). I tested a screen-recording-style clip to confirm: the embedded subtitle track was handled, the in-app UI strings were left alone. That is the right call — trying to repaint a moving software interface usually produces artefacts — but it means the 'video text' framing needs care: this tool handles added explanatory text, not every pixel of text in frame.

Third, on cost: at 10 points per minute (source: Vozo docs), a 20-minute video costs 200 points. Studio's 600-point monthly allowance therefore covers about 60 minutes of visual translation — the plan card says as much (source: vozo.ai/pricing). For occasional use the free 20 points are enough for a genuine trial (roughly 2 minutes, and the vendor's own sample videos cost nothing), and unused monthly points survive one extra month, which softened the batch-month problem for me. There is no pure pay-as-you-go mode; if you exceed your tier you buy point packs or change plans (source: Vozo docs FAQ).

What it does not do

The 20-minute per-file cap is the hard wall, on every tier including Enterprise — long-form content has to be cut down before it goes in (source: vozo.ai/pricing). Output is capped at 1080p even though input accepts 4K, so a 4K master gets downscaled in translation (source: vozo.ai/pricing). It does not translate speech, audio or lip movement — those are separate Vozo features with their own point costs (3 points per minute for dubbing, plus 1 for subtitle removal), and the unified workflow across visual text, subtitles, dubbing and lip sync is still listed as in progress rather than shipped (source: Vozo docs). It does not handle UI or real-world text, as above. And rebuilt text on heavily textured or fast-moving backgrounds is the weak case: the erase-and-repaint approach works on clean overlays and slides and gets less convincing as the background gets busier, which is inherent to the technique rather than a Vozo-specific flaw.

Verdict

Rated 3.7 on the rubric. Core functionality scores high for what it claims: the detect-translate-rebuild loop preserves styling and animation in a way subtitle tools cannot, the editor with per-element retranslate and timing control is genuinely production-grade, and in my testing the output on slide-and-overlay content was close to hand-localised quality. Value is fair — 10 points per minute is legible metering, the free trial is real, and Studio's 600 points cover about an hour of translation a month, though the 20-minute file cap and 1080p output ceiling pinch professional use. Usability is good: the side-by-side proofing view and one-element retranslate loop are exactly right for this task. Support and ecosystem are mid-pack: solid docs and an API on every paid plan, but the platform is small and the unified workflow is still promised rather than delivered. Reliability signals are decent — refunds on failed jobs, transparent point accounting, documented feature boundaries that match reality. Best for creators and L&D teams localising slide-heavy, overlay-heavy video — training, product demos, explainers — who need the on-screen text layer rebuilt, not just subtitled.

Sources

• https://www.vozo.ai/visual-translate
• https://www.vozo.ai/pricing
• https://www.vozo.ai/docs/visual_translate/get-started
• https://www.vozo.ai/docs/common/tools-points-rules
• https://www.vozo.ai/docs/api_reference/get_started
• https://www.vozo.ai/api
• https://www.vozo.ai/video-translate
• https://www.g2.com/products/vozo-ai/reviews
• https://www.vozo.ai/translate-video-to-english