Türkçe: README.tr.md
RemoteBite is a smartphone TV remote you can run entirely by voice. The user speaks naturally, for example “mute and go to channel 5 then open the guide” or “connect to my TV”, and the app understands the request, carries out each step on the TV and confirms by voice.
Everything happens on the phone. After a one-time download, voice control works fully offline: no audio leaves the device, no usage data goes to a server and nobody has to pay for a cloud service.
The end goal is “zero-click” control: every action in the app, from finding and connecting to a TV and waking it from standby to changing channels by name, changing settings and moving between screens, can be done by voice alone. A full touch remote sits alongside, so voice and touch cover the same ground.
A working prototype of the voice core exists. Zero-click control is the next milestone.
flowchart TD
U["User speaks naturally"] --> P["Phone understands the request on the device (offline, private)"]
P -->|unsure| Q["Asks: Did you mean...?"]
Q --> U
P -->|TV commands| T["TV carries out each step in order"]
P -->|app commands| Z["Zero-click: connect, switch TVs, change settings and navigate by voice, no touch"]
T --> C["Short spoken confirmation"]
Z --> C
R["Full touch remote alongside voice"] -.->|fallback| T
| Area | Examples |
|---|---|
| Volume | “Turn it up”, “Volume up 5 times”, “Set volume to 30”, “Mute”, “The volume is 35” |
| Channels | “Channel 42”, “Go to ESPN”, “Next channel”, “Go back to what I was watching”, “Show the channel list” |
| Playback | “Pause”, “Resume”, “Skip ahead 30 seconds”, “Record” |
| Power | “Turn off the TV”, “Turn on the TV”, “Set a sleep timer” |
| Navigation and keys | “Press OK”, “Open the guide”, “Go home”, “Press red” |
| Apps, inputs and subtitles | “Open Netflix”, “Switch to HDMI 2”, “Enable captions”, typing text on the TV |
| Connection | “Connect to my TV”, “Switch to the living room TV”, “Disconnect” |
| Moving around the app | “Go to settings”, “Show my TVs”, “Open the remote” |
| Settings | “Turn off haptic feedback”, “Change the language to Turkish”, “Set the wake word to hey remote” |
| Voice feature itself | “Download the AI model”, “Delete the speech model” |
Several requests can be combined in one sentence and run in order. A conversational mode lets the user talk back and forth with the assistant, and on iPhone the system’s voice shortcuts can trigger the same commands.
| Phase | Milestone |
|---|---|
| 1. Remote foundation | A complete touch remote: saved TVs, TV manager, wake from standby, three layouts, named inputs, settings, reliable reconnection and 18 interface languages. |
| 2. Private voice core | Basic TV commands by voice, fully on the phone, with spoken confirmations, learning from corrections, a conversational mode, iPhone voice shortcuts, and a fallback that keeps voice working on less capable phones. |
| 3. Better listening | More accurate recognition, a hands-free wake phrase, an optional fully offline speech recognizer and a visible “I’m hearing you” indicator. |
| 4. Voice for the whole app | Connection, extra keys, app navigation, settings and management of the voice feature itself, all reachable by voice, with confirmation for risky changes. |
| 5. Smarter conversations | Compound commands, awareness of the user’s saved TV and channel names, and follow-up questions for confirmation or clarification. |
| 6. Zero-click everywhere | Voice on every screen, connecting and waking TVs by voice, and optional listening on app launch. |
| 7. Knowing the TV’s state | Target volume (“volume to 20”), voice correction of the estimate and “last channel”. |
| 8. Real-world validation | Testing with real people, phones and TVs in several languages. |
The zero-click milestone is reached when, on real phones with a real TV, a user can: connect with no touches after setup, change channels by name, set a target volume, give compound commands, change settings and navigate by voice, wake the TV, still use voice on a less capable phone, and do all of this in airplane mode after the download.
| Stakeholder | Benefit |
|---|---|
| TV viewers | Natural-language control of the TV and the app, including compound and name-based requests, without hunting for buttons. |
| People who struggle with small buttons or with seeing them | Every action, from connecting to the TV to changing settings, can be done by voice, with spoken confirmation of what happened. |
| Privacy-conscious users | No audio or usage data leaves the phone. The assistant works offline. |
| Multilingual households | The interface in 18 languages, with voice commands in many of them. |
| Users with older or lower-end phones | Common voice commands stay available even where the full voice understanding cannot run. |
| App publisher | No server costs for voice, and a small app download because the voice feature is fetched after install. |
| Open-source and on-device AI community | A concrete example of private, offline AI voice control in an everyday consumer app. |
| Risk | Mitigation |
|---|---|
| Many older or lower-end phones cannot run the full voice understanding | Common voice commands still work on those phones instead of voice being switched off. |
| Understanding gets less reliable as users ask for more | The AI only interprets the request; the app itself decides the concrete steps, and uncertain requests get a follow-up question instead of a guess. |
| Misheard or unclear commands | The assistant asks “Did you mean…?”, says when it did not understand, and learns from the user’s corrections. |
| Risky changes made by mistake (language switch, deleting the voice feature, changing the wake phrase) | An explicit spoken confirmation before applying them. |
| The TV cannot report volume, channel or power state | Estimated state, voice correction and channel memory. The limitation is stated openly. |
| The one-time download fails or is too large for the connection or phone | Automatic retries, a storage check, Wi-Fi-only by default and clear spoken or on-screen errors. |
| Only Samsung TVs are supported in practice | The design leaves room for other brands; support for them is not built yet. |
The basic remote-control foundation is based on an MIT-licensed open-source Samsung TV remote project by mazen-salah; the ideas described here are the additions on top of it.
Keywords: voice control, smart TV, accessibility, offline AI, on-device AI, privacy, hands-free, assistive technology, remote control
If you find this idea useful, you can support the author’s open work through GitHub Sponsors.
Originally designed by Merih İlgör (2026); published here as an open project idea.
This document and the project idea it describes are licensed under CC BY-NC-SA 4.0 for non-commercial use (see LICENSE). Commercial use requires a revenue-share agreement (see COMMERCIAL-LICENSE.md). Third-party components, including the MIT-licensed remote-control base and the language and speech models, remain under their own licenses. Modified versions may not be sold or transferred to third parties as new ideas, and patent-like rights are reserved; see COMMERCIAL-LICENSE.md.