Preparing context
General chat may work, but the AI is not yet grounded in this video.
A LITTLE CURIOSITY. A LOT TO DISCOVER.
Your next video.
Your next aha!
Watch, ask, connect the dots. Get to know your AI video study companion—one small step at a time.
Let’s get startedAndroid phones & TVs · English guide · Updated November 8, 2026
WATCH▶
ASKWhy?
CONNECTAha! ✦
Made for curious minds.
01
GET READY
Set up your learning space
- Open Settings using the gear icon beside the app name. Choose English under Language if needed. The app also supports other interface languages.
- Select Sign in with Google. Choose your own Google account. Use the same TubePal account when restoring a purchase. Having Gmail on your phone alone does not sign you into TubePal.
- Under the Gemini API key setting, paste your key and save it. Never paste it into the conversation box, a review or a support email.
- Choose an available analysis model and Live model. These are separate settings: a text/video model name is not automatically a valid Live model.
- Connect headphones whenever possible. They prevent speaker audio from competing with your voice and let the video keep playing while you talk. Keep offline recognition and local text-to-speech off if you want Gemini Live only; Live still needs a network connection.
Your Gemini key is stored encrypted on the device. Provider quotas, supported regions and model access belong to your Google API project. TubePal does not promise unlimited free AI.
02
FOLLOW YOUR CURIOSITY
Open a video and let context catch up
- Tap the search icon. Enter keywords or paste a complete YouTube link such as
https://www.youtube.com/watch?v=VIDEO_ID(replace the placeholder with a real ID). - For a search, choose one of the result cards. For a link, open the matched video. The official YouTube player supplies playback controls and may serve its own ads.
- Allow the video context to finish preparing. Look for its status in the conversation area. Long videos and limited API quotas can take longer or fail.
- When the context is ready, ask about a particular moment. The app sends the video reference, available context and playback time to help the AI answer.
Context ready
Ask video-specific questions. Check the original moment when accuracy matters.
Context is AI-generated learning material, not a complete subtitle download. Some frames, spoken words, formulas or details may be missed. Public does not always mean embeddable or analyzable.
03
TALK IT THROUGH
Ask naturally. Keep your hands free.
Tap the vertical waveform beside Send to start voice conversation. Allow microphone access when Android asks. Wait for the Live connection status, then speak. Tap the waveform again to stop.
“Explain the idea at this point in simpler words.”
“Go forward 30 seconds.”
“Jump to 4 minutes.”
“Find the chapter about fermentation.”
An absolute timestamp (“jump to 4 minutes”) differs from a relative movement (“forward 4 minutes”). Topic jumps use the available chapters and context; the app may not find a match. Check the player position and confirmation before relying on the result.
- Prefer typing? Enter a question below the conversation and tap Send.
- Copy your notes: use the conversation copy control, or select text. It copies the currently displayed conversation, not all cloud history.
- Video too loud? In Settings, choose whether voice conversation pauses the video or lowers its volume. Use headphones to reduce feedback.
- Stay safe: do not operate machinery, drive, exercise beyond your limits or handle hot equipment based solely on AI instructions.
04
Read, export and revisit a moment
- Open Full subtitles. TubePal checks for an available YouTube caption track first. If none can be retrieved, Gemini can transcribe in two-minute sections. AI text may contain errors.
- Select a timestamp to replay the passage. Choose Translate for the interface language or return to the original. Export subtitles saves timestamped text for personal study.
- Open the mind map for linked topics or switch to the whiteboard for a walkthrough. These use available subtitles and visual context; check the video for details the AI may miss.
Originals and translations are cached locally. Signed-in users can reuse available shared Firebase study material. Unavailable cloud data does not invalidate a saved local copy.
05
MAKE IT YOURS
Keep good finds. Pick up where you left off.
- Open Favorites between Search and Settings. Save the current video, create a folder or move a favorite into a subfolder.
- Search the collection, share a video link or delete a saved entry. This app collection is not guaranteed to synchronize with your YouTube playlists.
- In Series playback settings, enable resuming and optionally set a default starting time for new viewing. The latest saved player position is used when it is available.
- Enter fullscreen for landscape viewing. Tap to reveal controls; use Back to return to portrait. Landscape keeps the video primary and offers compact voice/chapter controls.
A sudden force-stop or missing player telemetry may prevent the very last second from being saved. Switching languages does not mean you need to create a new account.
CLEAR BOUNDARIES
Your account, your key, your choices
Data & privacy
Starting Gemini Live transmits microphone audio to Gemini. Questions, answers, video references and timestamps may be processed by Gemini and synchronized to the authenticated project backend. Firebase stores supported account/history data. API keys are not copied into Firestore.
Use Settings to sign out or delete your account. For deletion without the app, visit the account-deletion page. Public, de-identified video knowledge may remain. See the policy for details.
Privacy policy · Delete accountFree BYOK release
This production release uses your own Gemini API key for AI features. Provider availability, free-tier limits and any provider charges apply separately. The app download does not include Gemini usage.
This build does not enable a TubePal paid checkout or TubePal advertising. It does not remove advertising served by YouTube. If monetization is introduced in a future update, the app and store listing will disclose it before use.
A LITTLE HELP
When something gets in the way
Live connects, but the AI cannot answer about my video
A Live audio connection and video context are different states. Wait for context readiness, then ask a question tied to the current timestamp. A generic answer does not prove video analysis succeeded.
Both the video and Live fail to connect
Check the phone’s actual access to Google and YouTube. A working Wi-Fi icon alone is not enough. Review your existing network/VPN setup without disabling security protections. Also check provider availability and the status shown in the app.
Analysis is slow, unavailable or rate-limited
Check the quota and model access for the Gemini project owning your key. A model listed in Settings may not be available to your project. Try an available supported model or a shorter embeddable video. Repeated retries can consume more quota. Error 429 commonly indicates a rate/quota limit; it does not prove the key is invalid.
I hear echoes or my voice is hard to recognize
Use headphones, verify microphone permission and check the selected audio mode. The app does not extract the YouTube audio track as an echo-cancellation reference, so perfect speaker echo removal is not guaranteed.
Google sign-in or purchase restore fails
Check Google Play services and try the intended Google account. The Play purchase and TubePal account must match. An unavailable purchase status does not mean you should buy again. Use Restore purchases and contact support if needed.
How do I send useful feedback?
Email support@gaoerji.net. Include the app version from About, phone model, Android version, approximate time/time zone, the exact error and steps to reproduce. Include a public video link only if you want to share it. Never send full API keys, passwords, purchase tokens, payment-card details or private chat.
中文快速使用说明
- 设置中登录自己的 Google 账号，填写并保存 Gemini Key；分析模型与 Live 语音模型分开选择。
- 点击放大镜输入关键词或 YouTube 完整网址，选择视频。等到“上下文就绪”再问视频内容；Live 连上不代表视频分析已完成。
- 点击输入框右侧的竖条波形开始说话，再点一次停止。建议戴耳机；视频暂停、原声降低可在设置调整。
- 打开完整字幕可按时间跳转；思维导图窗口含导图与白板。翻译尚未完成时可先看源语言。
- 收藏中可建子目录、移动和分享。连续剧设置中可配置起播时间与续播。
- 广告不打断视频，可直接关闭。一次性购买 4.99 美元的去广告升级只去除 APP 广告，不去除 YouTube 广告、不包含 Gemini 用量，以 Google Play 当地价格为准。
- 使用 API 图文申请指南，或联系 support@gaoerji.net。不要发送完整 Key、密码或支付令牌。
BIG SCREEN, SMALL REMOTE
TV, remote control & far-field microphones
Use the directional pad to move between framed controls, then press OK to activate. The main order is Chapters → Mind map → Search → Favorites → Settings → Player → Fullscreen → Conversation. On the framed player, OK toggles play/pause; Right enters supported native YouTube controls and Back returns to the player frame. Native controls vary by YouTube player version.
Frame the text input first, press OK to edit, and press Back to finish editing and return to the frame without sending. Use Up/Down to move to another control; select Send and press OK when ready.
A remote’s voice key may belong to the system assistant. It is not guaranteed to send microphone audio to TubePal. Where the system search integration is supported, choose the TubePal result card. For continuous AI conversation, select an available microphone in the app and activate the waveform.
A TV far-field microphone can work as an app input only if the TV exposes it to Android apps. A system notice that far-field assistance is off can mean the app has taken over recording. It does not prove an error. Loud video audio can still be mistaken for your voice: use headphones or the speaker pause/mute options. TubePal does not bypass device security or extract YouTube PCM for perfect echo cancellation.
HEAR EACH OTHER CLEARLY
Choose one speaker listening strategy
- Headphones first: keep video audio out of the room microphone. Bluetooth headset microphone availability and sound quality depend on Android, the headset and the active route; LE Audio is not guaranteed by the Bluetooth version alone.
- Pause video while listening on speakers: enabled by default. A connected session can remain stopped while video plays. Start listening with the waveform to pause the video. A playback/jump command that starts playback stops listening; video completion resumes listening.
- Mute video while listening on speakers: an alternative to pausing. Start listening to keep the video running silently. When a playback command resumes audible playback, listening stops; video completion resumes listening.
- Start listening when opening a video: on TV, enabling this disables both speaker pause and speaker mute. Those protections cannot operate together with that auto-listening choice.
- Lower original volume during conversation: choose a remaining video volume of 10%, 20%, 40%, 60% or 80%. This is volume reduction, not echo removal, and cannot make a paused video audible. A separate pause-during-conversation preference may also pause headphone playback.
- Offline recognition / local TTS: leave disabled for Gemini Live only. If disabled, a Live connection failure does not silently enable them.
SAY WHAT YOU WANT TO DO
Voice command reference
Speak after listening starts. Playback commands use a specific text confirmation without an additional spoken confirmation. A transcript alone does not prove the player accepted a command; check the confirmation and position.
- Playback: “Pause”, “Play”, “Replay”, “Go forward 30 seconds”, “Go back 2 minutes”.
- Exact time / topic: “Jump to 4 minutes”, “Go to the chapter about fermentation”. Topic matching needs available chapters or context and may request clarification.
- Loop: “Loop the previous 50 seconds”, “Loop from 5 to 6 minutes”, “Stop looping”. The range must fit the video.
- View: “Make the screen bigger”, “Return to portrait”, “Turn subtitles on”, “Turn subtitles off”. Subtitle controls require an available caption track.
- Speed / sound: “Play faster”, “Faster again”, “Play slower”, “Mute”, “Turn the sound on”. Available speeds depend on the player.
- Search / choose: “Search for beginner yoga”, then “Play the first video” (or second, third, fourth) from the current results.
- Favorites: “Save this video to favorites”, “Save this video to the Cooking folder”, “Open the pasta video from favorites”. Ambiguous matches may need a selection.
- Conversation: ask a question about the current moment, or say “Stop listening”. General chat can work before video context is ready, but is not grounded in the video yet.
Commands depend on recognized language, available player capabilities and model connectivity. Do not rely on voice control for safety-critical work.
ON A REAL PHONE
Watch it come together
Actual English screens from TubePal AI on a Vivo phone. Tap an image to inspect it. These are unretouched captures, not interface mockups.
Generated material can be incomplete or wrong. UNKNOWN marks unclear evidence; it is preserved rather than replaced with invented information. Check important claims against the video.
Demo film: Big Buck Bunny. (c) copyright 2008, Blender Foundation / www.bigbuckbunny.org. CC BY 3.0. No endorsement by Blender, YouTube or Google is implied. UI screenshots do not download or extract the film.
TubePal AI community · 官方社区
Discuss a video, share a learning tip or report a problem. 请勿公开 API Key。 Join TubePal AI on Discord · 加入官方频道