project: bump to 1.7.0

.github/workflows/build.yml

@@ -23,7 +23,7 @@ jobs:
     name: Build APK (Qualcomm DSP)
     runs-on: ubuntu-latest
     env:
-      ANDROID_APK_VERSION: 1.6.10
+      ANDROID_APK_VERSION: 1.7.0
 
     steps:
       - name: checkout code
@@ -121,7 +121,7 @@ jobs:
     name: Build APK (non-Qualcomm)
     runs-on: ubuntu-latest
     env:
-      ANDROID_APK_VERSION: 1.6.10
+      ANDROID_APK_VERSION: 1.7.0
       SKIP_HEXAGON_SDK: "1"
       SKIP_OPENCL_SDK: "1"
 

.gitignore

@@ -86,18 +86,17 @@ libggmldsp-skel*.so
 CMakeLists.txt
 
 prebuilts/
-core/ggml/llamacpp/ggml/src/ggml-opencl/
-core/ggml/llamacpp/ggml/src/ggml-sycl/
-core/ggml/llamacpp/ggml/src/ggml-virtgpu/
-core/ggml/llamacpp/ggml/src/ggml-webgpu/
-core/ggml/llamacpp/ggml/src/ggml-zdnn/
-core/ggml/llamacpp/ggml/src/ggml-zendnn/
-core/ggml/llamacpp/ggml/src/ggml-vulkan/
-core/ggml/llamacpp/ggml/src/ggml-openvino/
-core/ggml/llamacpp/ggml/src/ggml-sycl/
-core/ggml/llamacpp/ggml/src/ggml-rpc/
-core/ggml/llamacpp.bak.20260724150435/
-core/ggml/llamacpp/ggml/src/ggml-hexagon/htp/ggml_dsp.h
-core/ggml/llamacpp/ggml/src/ggml-hexagon/htp/ggml_dsp_skel.c
-core/ggml/llamacpp/ggml/src/ggml-hexagon/htp/ggml_dsp_stub.c
+core/llamacpp/ggml/src/ggml-opencl/
+core/llamacpp/ggml/src/ggml-sycl/
+core/llamacpp/ggml/src/ggml-virtgpu/
+core/llamacpp/ggml/src/ggml-webgpu/
+core/llamacpp/ggml/src/ggml-zdnn/
+core/llamacpp/ggml/src/ggml-zendnn/
+core/llamacpp/ggml/src/ggml-vulkan/
+core/llamacpp/ggml/src/ggml-openvino/
+core/llamacpp/ggml/src/ggml-sycl/
+core/llamacpp/ggml/src/ggml-rpc/
+ggml_dsp.h
+ggml_dsp_skel.c
+ggml_dsp_stub.c
 log*.txt

README.md

@@ -1,29 +1,40 @@
 # KanTV
 
-KanTV("Kan", aka English "watch") , an open source project focus on study and practise on-device AI technology in <b>real scenario</b>(such as perform <b>online-TV playback</b> and <b>realtime transcription</b> and <b>online-TV record</b> at the same time) on Android phone:
+KanTV ("Kan", meaning "watch" in English and "看" in Chinese), an open source project focusing on studying and practicing on-device AI technology in <b>real scenarios</b> (such as performing <b>online-TV playback</b>, <b>realtime transcription</b>, and <b>online-TV recording</b> at the same time) on Android phones:
 
 
-- Watch online TV and local media by customized ![FFmpeg 6.1](https://github.com/zhouwg/FFmpeg). this project is derived from original ![ijkplayer](https://github.com/zhouwg/kantv/tree/kantv-initial)(that project has stopped maintenance since 2021), with much enhancements and new features. source code of customized FFmpeg 6.1 could be found in <a href="https://github.com/zhouwg/kantv/tree/master/external/ffmpeg-6.1"> external/ffmpeg </a>according to <a href="https://ffmpeg.org/legal.html">FFmpeg's license</a>. source code of FFmpeg 6.1's all dependent libraries could be found in <a href="https://github.com/zhouwg/kantv/tree/master/external/ffmpeg-deps"> external/ffmpeg-deps </a>.
+- Watch online TV and local media using a customized ![FFmpeg 6.1](https://github.com/zhouwg/FFmpeg). This project is derived from the original ![ijkplayer](https://github.com/zhouwg/kantv/tree/kantv-initial) (that project has stopped maintenance since 2021), with many enhancements and new features. Source code of the customized FFmpeg 6.1 can be found in <a href="https://github.com/zhouwg/kantv/tree/master/external/ffmpeg-6.1"> external/ffmpeg </a> according to <a href="https://ffmpeg.org/legal.html">FFmpeg's license</a>. Source code of all FFmpeg 6.1's dependent libraries can be found in <a href="https://github.com/zhouwg/kantv/tree/master/external/ffmpeg-deps"> external/ffmpeg-deps </a>.
 
-- Watch online TV by customized ![Google Exoplayer 2.15.1](https://github.com/google/ExoPlayer), source code of customized Exoplayer2.15.1 could be found in <a href="https://github.com/zhouwg/kantv/tree/master/android/kantvplayer-exo2"> android/kantvplayer-exo2 </a>.
+- Watch online TV using a customized ![Google ExoPlayer 2.15.1](https://github.com/google/ExoPlayer). Source code of the customized ExoPlayer 2.15.1 can be found in <a href="https://github.com/zhouwg/kantv/tree/master/android/kantvplayer-exo2"> android/kantvplayer-exo2 </a>.
 
-- Record online TV to local file on phone.
+- Record online TV to a local file on the phone.
 
 - 2D graphic performance benchmark.
 
-- AI subtitle(real-time English subtitle for English online-TV(aka OTT TV) via the great & excellent & amazing<a href="https://github.com/ggerganov/whisper.cpp"> whisper.cpp </a>).
+- AI subtitle (real-time English subtitle for English online-TV (aka OTT TV) via the great & excellent & amazing <a href="https://github.com/ggerganov/whisper.cpp"> whisper.cpp </a>).
 
-- Well-maintained <b>turn-key / self-contained</b> workbench for AI experts/researchers whom focus on highly-value on-device AI R&D activity on Android. some on-device AI R&D activities (AI algorithm validation and AI model validation and performance benchmark with ASR/LLM/MTMD(multimodal) on Android) could be done via this project easily.
+- Well-maintained <b>turn-key / self-contained</b> workbench for AI experts/researchers who focus on high-value on-device AI R&D activities on Android. Some on-device AI R&D activities (AI algorithm validation, AI model validation, and performance benchmark with ASR/LLM/MTMD (multimodal) on Android) can be done via this project easily.
 
-- Well-maintained <b>turn-key / self-contained</b> workbench for AI beginners to learning on-device AI technology on Android.
+- Well-maintained <b>turn-key / self-contained</b> workbench for AI beginners to learn on-device AI technology on Android.
 
-- Built-in [Alibaba's Qwen1.5-1.8B](https://huggingface.co/Qwen/Qwen1.5-1.8B-Chat-GGUF), [Alibaba's Qwen2.5-3B](https://huggingface.co/Qwen/Qwen2.5-3B-Instruct-GGUF), [Alibaba's Qwen3-4B](https://huggingface.co/Qwen/Qwen3-4B/tree/main), [Google's Gemma3-4B(multimodal text + image)](https://huggingface.co/ggml-org/gemma-3-4b-it-GGUF/tree/main), [Google's Gemma-4-E2B](https://huggingface.co/unsloth/gemma-4-E2B-it-GGUF), [Huggingface's SmolVLM2-256M(highly-optimized multimodal for realtime-video-recognition)](https://huggingface.co/ggml-org/SmolVLM2-256M-Video-Instruct-GGUF), [Alibaba's Qwen2.5-Omni-3B(multimodal text + audio)](https://huggingface.co/ggml-org/Qwen2.5-Omni-3B-GGUF) supportive and runs entirely <b>offline(no Internet required)</b>. these supported LLM models can be [downloaded in the Android APK directly](./docs/how-to-download-ai-models.md) without manually preparation. APK's users can compare the <b>real experience</b> of these LLM models on the Android phone. developers can add other LLM models manually in source code [KANTVAIModelMgr.java](https://github.com/zhouwg/kantv/blob/master/android/kantvplayer-lib/src/main/java/kantvai/ai/KANTVAIModelMgr.java).
+- Built-in AI models are supported and run entirely <b>offline (no Internet required)</b>. These supported AI models can be [downloaded in the Android APK directly](./docs/how-to-download-ai-models.md) without manual preparation. APK users can compare the <b>real experience</b> of these AI models on the Android phone. Developers can add other AI models manually in the source code [KANTVAIModelMgr.java](https://github.com/zhouwg/kantv/blob/master/android/kantvplayer-lib/src/main/java/kantvai/ai/KANTVAIModelMgr.java).
 
-- The [ggml-hexagon](https://github.com/zhouwg/kantv/blob/master/core/ggml/llamacpp/ggml/src/ggml-hexagon/ggml-hexagon-jz.cpp) in this project is probably the first open-source reference implementation of a specified llama.cpp backend for Qualcomm Hexagon NPU on Android phone. The backend type (Hexagon cDSP vs. generic ggml) is decided at build time, and the DSP-side thread count is automatically clamped based on the target SoC (e.g., 6 threads on Snapdragon 8Elite, 4 threads on Snapdragon 8Gen3).
+  | Model | Type | Capability | Source |
+  |-------|------|-----------|--------|
+  | ggml-tiny.en-q8_0 | ASR | speech-to-text | [whisper.cpp](https://github.com/ggerganov/whisper.cpp) |
+  | [Qwen1.5-1.8B](https://huggingface.co/Qwen/Qwen1.5-1.8B-Chat-GGUF) | LLM | text-only | Alibaba |
+  | [Qwen2.5-3B](https://huggingface.co/Qwen/Qwen2.5-3B-Instruct-GGUF) | LLM | text-only | Alibaba |
+  | [Gemma3-4B](https://huggingface.co/ggml-org/gemma-3-4b-it-GGUF/tree/main) | LLM | text + image (MTMD) (known issue: garbled output) | Google |
+  | [Gemma-4-E2B](https://huggingface.co/unsloth/gemma-4-E2B-it-GGUF) | LLM | text-only (default) | Google |
+  | [SmolVLM2-256M](https://huggingface.co/ggml-org/SmolVLM2-256M-Video-Instruct-GGUF) | LLM | text + image (realtime-video-recognition) (known issue: regression after merging upstream code, video inference broken, only video preview works, WIP) | Huggingface |
+  | [Qwen2.5-Omni-3B](https://huggingface.co/ggml-org/Qwen2.5-Omni-3B-GGUF) | LLM | text + audio (MTMD) | Alibaba |
+
+- The [JZ's ggml-hexagon](https://github.com/zhouwg/ggml-hexagon) used in this project is probably the first open-source reference implementation of a specific llama.cpp backend for Qualcomm Hexagon NPU on Android phones. The backend type (Hexagon cDSP vs. generic ggml) is decided at build time, and the DSP-side thread count is automatically clamped based on the target SoC (e.g., 6 threads on Snapdragon 8Elite, 4 threads on Snapdragon 8Gen3). Its PP (prompt processing) and TG (token generation) performance comprehensively surpasses [Qualcomm's official implementation](https://github.com/ggml-org/llama.cpp/tree/master/ggml/src/ggml-hexagon) on Snapdragon 8Elite (aka 8Gen4); benchmark comparisons can be found [here](https://github.com/zhouwg/ggml-hexagon/blob/self-build-jz/docs/backend/jz-ggml-hexagon/ion-mempool-vs-perbuffer-analysis-20260713.md).
 
 ### Software architecture of KanTV Android
 
-![Image](https://github.com/user-attachments/assets/32227392-abdf-4760-be3d-a19d48d478ff)
+<img width="803" height="517" alt="kantv-arch" src="https://github.com/user-attachments/assets/59e62514-0c1d-44fb-a4fe-dd8d805e17cb" />
+
 
 ### Building the project
 
@@ -32,61 +43,34 @@ KanTV("Kan", aka English "watch") , an open source project focus on study and pr
 - Download pre-built Android APK from Github CI-build: https://github.com/zhouwg/kantv/actions/
 
 ### Run Android APK on Android phone
-- Android 8.0(2017.08) --- Android 15(2024.10) and higher version with <b>ANY</b> mainstream arm64 mobile SoC.
-- Android smartphone equipped with <b>ANY</b> mainstream <b>high-end</b> mobile SoC is highly <b>recommented</b> for realtime AI-subtitle feature otherwise unexpected behavior would happen.
-- Android smartphone equipped with one of below Qualcomm mobile SoCs is required for verify/running ggml-hexagon backend on Android phone:
+- Android 8.0 (2017.08) --- Android 15 (2024.10) and higher versions with <b>ANY</b> mainstream arm64 mobile SoC.
+- An Android smartphone equipped with <b>ANY</b> mainstream <b>high-end</b> mobile SoC is highly <b>recommended</b> for the realtime AI-subtitle feature, otherwise unexpected behavior may occur.
+- An Android smartphone equipped with one of the below Qualcomm mobile SoCs is required for verifying/running the ggml-hexagon backend on Android phones:
 ```
     Snapdragon 8 Gen 2
     Snapdragon 8 Gen 3
-    Snapdragon 8 Elite(aka 8 Gen 4)
-    Snapdragon 8 Elite Gen 5(aka 8 Gen 5)
+    Snapdragon 8 Elite (aka 8 Gen 4)
+    Snapdragon 8 Elite Gen 5 (aka 8 Gen 5)
 ```
 
 
 ### Screenshots
 
-here is a short video to demostrate realtime AI subtitle by running the great & excellent & amazing<a href="https://github.com/ggerganov/whisper.cpp"> whisper.cpp </a> on an Android phone equipped with Qualcomm Snapdragon 8Gen3 mobile SoC - <b>fully offline, on-device</b>.
+Here is a short video to demonstrate realtime AI subtitle by running the great & excellent & amazing <a href="https://github.com/ggerganov/whisper.cpp"> whisper.cpp </a> on an Android phone equipped with Qualcomm Snapdragon 8Gen3 mobile SoC - <b>fully offline, on-device</b>.
 
 https://github.com/zhouwg/kantv/assets/6889919/2fabcb24-c00b-4289-a06e-05b98ecd22b8
 
 ----
 
-a screenshot to demostrate multi-modal inference by running the magic <a href="https://github.com/ggerganov/llama.cpp"> llama.cpp </a> on an Android phone equipped with Qualcomm Snapdragon 8Elite mobile SoC  - <b>fully offline, on-device</b>.
+A screenshot to demonstrate multi-modal inference by running the magic <a href="https://github.com/ggerganov/llama.cpp"> llama.cpp </a> on an Android phone equipped with Qualcomm Snapdragon 8Elite mobile SoC  - <b>fully offline, on-device</b>.
 
 ![Image](https://github.com/user-attachments/assets/c406951a-383a-4943-a58d-cda401148f9e)
 
 ----
-a screenshot to demostrate realtime-video-recognition via [MTMD from llama.cpp](https://github.com/ggml-org/llama.cpp/blob/master/docs/multimodal.md) + a lightweight multimodal model [SmolVLM2-256M from Huggingface](https://huggingface.co/HuggingFaceTB/SmolVLM2-256M-Video-Instruct) on an Android phone equipped with Qualcomm Snapdragon 8Elite mobile SoC  - <b>fully offline, on-device</b>.
+A screenshot to demonstrate realtime-video-recognition via [MTMD from llama.cpp](https://github.com/ggml-org/llama.cpp/blob/master/docs/multimodal.md) + a lightweight multimodal model [SmolVLM2-256M from Huggingface](https://huggingface.co/HuggingFaceTB/SmolVLM2-256M-Video-Instruct) on an Android phone equipped with Qualcomm Snapdragon 8Elite mobile SoC  - <b>fully offline, on-device</b>.
 
 ![Image](https://github.com/user-attachments/assets/35841e4d-150f-4163-bc58-ada1e9b1a065)
 
-<details>
-  <summary>some other screenshots</summary>
-  <ol>
-
-![Image](https://github.com/user-attachments/assets/d9c9bc39-d0d8-4d50-b74d-59152de28d6d)
-
-![Image](https://github.com/user-attachments/assets/025a8ff0-7584-4df2-97a5-f4e655a52e0f)
-
-
-----
-
-a screenshot to demostrate ASR inference by running the excellent <a href="https://github.com/ggerganov/whisper.cpp"> whisper.cpp </a> on an Android phone equipped with Qualcomm Snapdragon 8Gen3 mobile SoC - <b>fully offline, on-device</b>.
-
-![Image](https://github.com/user-attachments/assets/46856bf2-cc4b-4b0a-9209-d07825fba2e7)
-
-
-----
-a screenshot to demostrate download LLM model in APK.
-
-![1213951738](https://github.com/user-attachments/assets/5a0a965e-1752-475e-a2c1-63e6f60a9009)
-![1242080159](https://github.com/user-attachments/assets/32586234-4b2c-4d43-b0ab-498c56de44b3)
-
-  </ol>
-</details>
-
-
-
 
 ### Docs
 - [How to build](./docs/how-to-build.md)
@@ -97,17 +81,16 @@ a screenshot to demostrate download LLM model in APK.
 - [ChangeLog](./release/README.md)
 - [ggml-hexagon:history of ggml-hexagon](https://github.com/zhouwg/ggml-hexagon/discussions/18)
 - [ggml-hexagon:high-level data path of ggml-hexagon](https://github.com/zhouwg/ggml-hexagon/discussions/33)
-- [Roadmap](https://github.com/zhouwg/kantv/discussions/262)
 
 
 ### Contribution
 
-Report issue in Android phone equipped with <b>mainstream</b> mobile SoC or submit PR to this project is greatly welcomed.
+Reporting issues on Android phones equipped with <b>mainstream</b> mobile SoCs or submitting PRs to this project is greatly appreciated.
 
-I use [GitHub issues](https://github.com/zhouwg/kantv/issues) for tracking feature requests and issue reports, please see [how to submit issue in this project ](https://github.com/zhouwg/kantv/issues/1).
+I use [GitHub issues](https://github.com/zhouwg/kantv/issues) for tracking feature requests and issue reports, please see [how to submit an issue in this project](https://github.com/zhouwg/kantv/issues/1).
 
 <!--
-comment out this section because some contributors in the upstream project might-be don't want to be appeared here
+comment out this section because some contributors in the upstream project might not want to appear here
 
 ### Contributors
 
@@ -118,37 +101,8 @@ comment out this section because some contributors in the upstream project might
 
 ### Special Acknowledgement
 
- <ul>AI inference framework
-
-   <ul>
-  <li>
-   <a href="https://github.com/ggml-org/ggml">GGML</a>
-  </li>
-
-
-  </ul>
-
-  </ul>
-
- <ul>AI application engine
-
-  <ul>
-  <li>
-   ASR engine <a href="https://github.com/ggml-org/whisper.cpp">whisper.cpp</a>
-  </li>
-
-  <li>
-   LLM engine <a href="https://github.com/ggml-org/llama.cpp">llama.cpp</a>
-  </li>
-
-  <li>
-   CV engine <a href="https://github.com/nihui/opencv-mobile">opencv-mobile</a>
-  </li>
-
-  <li>
-   MTMD(multimodal) engine <a href="https://github.com/ggml-org/llama.cpp/blob/master/tools/mtmd/README.md">MTMD subsystem in llama.cpp</a>
-  </li>
-
-  </ul>
-
-  </ul>
+- Inference engine [GGML](https://github.com/ggml-org/ggml)
+- ASR engine [whisper.cpp](https://github.com/ggml-org/whisper.cpp)
+- LLM engine [llama.cpp](https://github.com/ggml-org/llama.cpp)
+- CV engine [opencv-mobile](https://github.com/nihui/opencv-mobile)
+- MTMD (multimodal) engine [MTMD subsystem in llama.cpp](https://github.com/ggml-org/llama.cpp/blob/master/tools/mtmd/README.md)

android/constants.gradle

@@ -28,8 +28,8 @@ project.ext {
     appTargetSdkVersion = 29
 
     //version and version code of KanTV Android APP
-    releaseVersion     = '1.6.10'
-    releaseVersionCode = 1000610
+    releaseVersion     = '1.7.0'
+    releaseVersionCode = 1007000
 
     //version and version code of internal customized Exoplayer2
     releaseVersionExoplayer     = '2.15.1'

android/gradle.properties

@@ -3,7 +3,7 @@ GROUP=KanTV.AI
 
 # http://central.sonatype.org/pages/requirements.html
 POM_USER_ORG=KanTV.AI
-POM_DESCRIPTION=workbench for study and practise on-device AI technology in real scenario with online-TV on Android phone, powered by llama.cpp & whisper.cpp & stable-diffusion.cpp and FFmpeg and original made ggml-hexagon
+POM_DESCRIPTION=workbench for study and practise on-device AI technology in real scenario with online-TV on Android phone, powered by llama.cpp & whisper.cpp and FFmpeg and original made ggml-hexagon
 POM_URL=https://github.com/zhouwg/kantv
 POM_SCM_URL=https://github.com/zhouwg/kantv
 POM_LICENSE_NAME=MIT license

android/kantvplayer-lib/src/main/java/kantvai/ai/KANTVAIModelMgr.java

@@ -200,7 +200,7 @@ private void initAIModels() {
          }
 
          try {
-             KANTVLibraryLoader.load("ggml-jni");
+             KANTVLibraryLoader.load("kantv-ai");
              KANTVLog.g(TAG, "cpu core counts:" + ggmljava.get_cpu_core_counts());
          } catch (Exception e) {
              KANTVLog.g(TAG, "failed to initialize ggml jni");