What can this video solve?
You will complete preparations for the first AI follow-up, and learn about the performance of normal follow-up, waiting and unmatched, as well as the uniform fallback when AI cannot be used.
Spoken and scripted
00:00-00:35 | Switch to AI Follow
Screen: Enter the full-screen prompt, tap the top mode switch, and switch from "Constant Speed" to "AI"; display prompts for logging in and starting a free trial. Spoken reading: AI Follow will move the text according to your reading progress. After entering the prompt page, tap the mode switch at the top to switch to AI. If you have not logged in yet, please follow the page prompts to complete the login and start the free trial; this step requires an Internet connection. The app then requests microphone permission, which is used to locally identify the location you read.
00:35-01:15 | Prepare the on-device model
Screen: Show model loading, initialization, warm-up, and a real-time speed check. Narration: After free access is enabled, the app checks the on-device LARGE model, initializes and warms up the recognition engine, then tests real-time performance. Preparation and recognition happen on the device, and scripts or audio are not uploaded. The first preparation may take longer, so keep this page open.
01:15-01:55 | Start reading
Screen: Show AI Ready, start prompting, read at a natural pace, then display the active words and movement. Narration: When AI Ready appears, start prompting and read naturally from the visible text. After Tdx Teleprompter confirms a match, the script moves and the prompt color marks the recognized position. Chinese, English, and mixed-language scripts are supported.
01:55-02:35 | Waiting and unmatched speech
Screen: Stop speaking to show Waiting to Speak, keep the script in place when speech cannot be matched, then resume from the visible text. Narration: Waiting to Speak is normal when you pause. If the recognized speech cannot be matched reliably, the script stays in place while AI Follow keeps listening. Continue reading from the visible text and following resumes after a stable match.
02:35-03:05 | Adjust position manually
Screen: Drag to another paragraph, release, and start reading from the new position. Narration: After a manual jump, AI Follow waits for speech from the newly visible text instead of returning to the old result. Read a short section near the focus line to help it find the new position.
03:05-03:35 | Floating AI Follow
Screen: Show the floating AI permission guide. Narration: Full-screen AI Follow requires microphone permission, but not floating-window or accessibility permission. To place AI Follow over the system camera, also enable the Tdx Teleprompter accessibility service. It does not read other app content or simulate clicks. It only supports the floating prompt and its audio connection.
03:35-04:00 | Return to fixed speed
Screen: Show the AI preparation error and Return to Fixed Speed. Narration: If device performance, memory, model files, or recognition components do not meet the requirements, the page explains why. You can return to fixed-speed mode and continue prompting. If failures repeat, select the relevant AI session in Feedback and share the diagnostic package through the system share sheet.
Offline text steps
- Enter the full-screen or Floating Prompt page, and click the top mode switch to switch to AI.
- As prompted on the page, complete the login and start the free trial; this step requires an Internet connection.
- Press boot to allow microphone permission.
- Wait for offline model checking, initialization, warm-up, and speed testing to complete on first use.
- Click to start prompting and read naturally from near the currently visible text.
- When "Waiting to speak" appears or the text temporarily stops, return to the currently visible original and read a short paragraph continuously.
- After manual scrolling, read again from near the focus line and wait for the AI to find its position.
- Floating AI Follow also requires the Tdx Teleprompter accessibility service.
- When the AI cannot prepare, it returns to the constant speed mode, which does not affect the basic prompting.