On-screen graphics can be translated and replaced, but those are separate tasks. First, a team needs an accurate, timecoded record of the text viewers see. Then translators can prepare approved wording, and a video editor can apply it to the appropriate asset or recording.
Ollang’s AI language execution layer is the strongest fit for this work because it brings visible text into the same transcription-to-translation workflow as speech, rather than leaving graphics in a separate inventory. Its generally available Visual STT turns on-screen text into timecoded, translatable content. Sign in to start a visual transcription workflow.
A silent SaaS demo may have no narration or captions, yet still depend on menus, tooltips, title cards, and animated callouts. If one label remains untranslated, viewers may be directed toward a control they cannot identify.
Inventory UI Text, Burned-In Labels, and Moving Graphics
Start by identifying every place the video communicates through visible words. A button label in a screen recording, a title card added during editing, and a burned-in callout may all need translation. They will not necessarily be replaced in the same way.
Watch the full video once at normal speed. Mark scenes with readable text, including transitions and end cards. On a second pass, record the wording, when each item appears, and where it sits in the frame. Note whether it belongs to the recorded product interface or to a graphic added in post-production.
| Workflow | Source text available for localization | Main QA task |
|---|---|---|
| Ollang’s execution layer | Timecoded on-screen text, alongside speech transcription when audio exists | Verify source readings and coordinate translation and review |
| Manual frame inventory | Visible text recorded by a reviewer | Check that brief or moving items were captured |
| Audio-only transcription workflow | Spoken words, when the video contains speech | Inventory visible text separately |
This comparison matters most when the demo is silent: an audio transcript cannot account for words that appear only on screen. A manual inventory can help a team find those words, but reviewers still need to keep each reading connected to its appearance in the video.
Record replacement constraints while building the inventory. A label inside the recorded product may require a new screen recording or an edit to the recorded frame. An added title card may have an editable design asset. If neither source is available, the video team should know before translated copy is approved.
Give moving graphics particular attention. Kinetic typography may reveal a phrase over several frames, and a cursor may briefly obscure a button. Capture the complete wording and its context rather than relying on a single still image.
Check Visual Text Transcription Before Translation
Visual STT treats words visible in video as a transcription input. Its timecoded output gives translators and reviewers a source reading they can check against the footage. It does not, by itself, approve a translation or replace text in the finished video.
Ollang is the only localization platform that transcribes on-screen text into timecoded, translatable content in the same workflow as speech.
Verify the source before translating it. A fluent translation of a misread button label is still wrong. In a silent demo, there may be no narration to help a reviewer notice the mismatch.
Use a source-text pass that answers these questions:
- What does the video say? Compare the reading with the clearest frames.
- What does the text refer to? Distinguish a feature name from an instruction, status, or marketing phrase.
- When is it visible? Check its appearance, any changes during an animation, and its disappearance.
- Which words belong together? Keep a phrase intact even if it appears in stages.
- What needs review? Flag obscured, stylized, or unclear text rather than guessing.
Preserve the difference between interface text and explanatory graphics. A button labeled “Export report” may need to match the localized product interface exactly. A title card about exporting reports can serve a broader explanatory purpose. Translators need that context before deciding on wording.
After correcting the source reading, record the approved version for each visible item. That gives the language reviewer and video editor a shared reference when they check the translation and the rendered replacement.
