This page covers how to choose the presenter image, write and record scripts that sound natural, and keep a series looking consistent.
Choosing the face of your AI spokesperson video
Use a well-lit, half-body photo with a neutral or branded background and relaxed posture. Business attire or brand colours help the clip feel intentional. If the spokesperson is a real employee, get written consent that covers synthetic video, how long it will be used and where.
Horizontal crops suit website heroes, help centres and presentations; vertical crops suit LinkedIn and short-form social. Prepare both from the same photo session.
Scripts that sound human
Presenter clips feel synthetic when the script reads like a brochure. Instead:
- Write how people talk: short sentences, contractions, one idea per line.
- Open with the benefit, not the company history.
- Keep explainers under 90 seconds; split longer topics into a series.
- Record with natural pauses — the model uses them for nods and breaths.
Formats that work
Product announcements, feature walkthroughs to pair with screen recordings, onboarding welcomes, internal updates, course introductions and FAQ answers all fit the format. Because videos can run up to five minutes in a single take, a complete lesson module doesn’t need to be stitched from fragments.
Localisation is straightforward: record or dub the script in another language and render it with the same presenter image.
Keeping a series consistent
Reuse one master photo, keep the same framing and record in the same room with the same microphone. Use a light, repeated direction prompt such as “confident, open hand gestures, static medium shot” so every episode has the same presence.
Approvals and compliance
Treat synthetic presenter videos like any other brand asset. Get legal sign-off on the consent form, keep a record of which scripts were approved, and avoid regulated claims — financial returns, health outcomes, guarantees — unless your compliance team has reviewed them. Labelling the video as AI-generated where platforms or regulations require it protects both the company and the person whose likeness is used.
Measuring what works
Track completion rate and click-through for each video rather than views alone. Short explainers with the key point in the first five seconds usually win; if completion drops sharply at one moment, rewrite and re-render just that script.
Localising with one presenter
For each market, translate and adapt the script, record or synthesise it in the target language and render it with the same presenter photo. Lip movement follows the new audio, so the result looks native rather than dubbed over the original performance.
FAQ
Can I use a photo of our CEO?
Yes, with their explicit consent. Agree in writing on where and how long the synthetic videos will be used.
Do I need text-to-speech?
No. A real recording usually sounds best, but you can upload TTS audio if you prefer.
How long can a presenter video be?
Up to 5 minutes per render, matching your audio.
Can I add slides or screen recordings?
Render the presenter, then combine it with slides or screen captures in your video editor.
Is there a watermark?
No. Paid renders download as clean MP4 files.
What resolution is the video?
720p, suitable for websites, LMS platforms and social media.
How do I get a consistent voice across videos?
Record all scripts with the same microphone in the same room, or use one fixed voice in your text-to-speech tool.
Can I update a video when pricing or features change?
Yes. Edit the script, record the new audio and render again with the same presenter photo — no reshoot needed.
Can the presenter appear next to a screen recording?
Render the presenter on a plain background, then place it as a picture-in-picture or side panel over your screen capture in an editor.