WeChat Keyboard launches new AI Voice Recorder feature! Real-time transcription and webpage generation. 3DM2026-10-2705:03:28 Today, Tencent announced the launch of the AI Voice Recorder feature for WeChat Keyboard. In terms of usage, mobile users can access the AI Voice Recorder through the WeChat Keyboard toolbar or directly within the app. PC users can invoke the tool panel using the shortcut key V, or access it via the Mac top menu bar or Windows toolbar. According to reports, the AI Voice Recorder can transcribe recordings in real-time into clearly structured webpages. It not only automatically extracts key points and organizes logic but also generates charts, diagrams, and interactive pages based on the content. WeChat Keyboard's AI Voice Recorder is not just a simple speech-to-text tool. It accurately recognizes speech content, removes typos and redundant expressions, and corrects content and organizes logic based on context without altering the original meaning, making scattered spoken content clearer and more organized. For example, after a trip, if you want to share your travel guide with friends, you can simply dictate the relevant content. The AI Voice Recorder will automatically organize it into a clearly structured webpage, which can then be shared to WeChat group chats with one click. This feature can also be used in daily life scenarios, such as dictating recipes, organizing rental precautions, or leaving instructions for pet sitting, allowing content that would normally require manual organization to be quickly completed via voice. For content with large amounts of information or scattered expressions, the AI Voice Recorder can reorganize the layout based on key points and emphasis, resulting in well-structured output that is easy to read and use. For instance, teachers can directly dictate teaching objectives, knowledge points, example problems, and common mistakes. The AI Voice Recorder can then organize this content into mind maps, courseware frameworks, or initial drafts of teaching pages, reducing the workload of manual organization. Additionally, the AI Voice Recorder has capabilities to assist with comprehension. For content such as open classes, professional lectures, technical live streams, or knowledge-based podcasts, it not only records and organizes notes but also extracts relationships between different concepts. It can generate charts, diagrams, short animations, or interactive pages based on the lecture content, making abstract and complex knowledge more intuitive.

Tencent open-sources AI model for real-time Chinese-English interpretation. Tencent has released Hy-SimulMT, an open-source model for simultaneous Chinese-English interpretation. It processes incoming speech continuously and supports translated text and speech output, with real-time voice cloning designed to preserve the original speaker's vocal characteristics. Users can also try the technology in Tencent's Hy Translate app. The release includes model weights, inference software and training [...] By TechNode Got news to share with the events industry? More in Technology & Digital Transformation. Related insights. Relevant events. Events in Technology & Digital Transformation

Tencent opens Hy-Nomas, an interactive video world model running on one RTX 4090. Tencent's Hunyuan team released Hy-Nomas, an open source interactive video world model with text and camera control that runs on a single RTX 4090. · 1 day ago Read 5 min * Tencent Hunyuan released Hy-Nomas, an open-source interactive video world model under Apache 2.0. * Supports free-form text prompts and camera trajectory control across autoregressive video segments. * A lifelong KV cache preserves world state and object consistency across long multi-turn generations. * Runs on a single RTX 4090 via INT8 DiT weights, sliding-window history, and CPU offload. * Ships cinematic and dual-arm robotics LoRA adapters, plus full fine-tuning and captioning pipelines. * Weights, demo, and training code live on Hugging Face and the web demo. Tencent's Hunyuan team has released Hy-Nomas, an Apache 2.0-licensed interactive video world model. It generates scenes one segment at a time, accepting new free-form text instructions or camera trajectories during generation. The release includes training code, inference scripts, a web demo, LoRA fine-tuning tools, and Hugging Face weights, including an INT8 checkpoint configured for a single NVIDIA GeForce RTX 4090. In Hy-Nomas, "world model" refers to a video generator that predicts each new segment from prior generated footage and the latest control input. Users can redirect the scene during a session while the model carries visual context forward, creating a development loop closer to an interactive environment than a fixed prompt-to-video clip. How Hy-Nomas remembers a scene. Hy-Nomas uses a diffusion transformer, or DiT, to generate video in autoregressive segments. Diffusion handles iterative denoising, while the transformer processes the sequence and its controls. Each segment receives either a text prompt or a camera-control signal. The model's "lifelong KV cache" stores attention keys and values from earlier segments. Giving later segments access to that history helps preserve objects, lighting, and geometry as generation continues, although the release does not yet provide benchmarks for long-horizon consistency. * Fast sampling: The reference interactive setup uses eight denoising steps. The distilled checkpoint also supports four-step and six-step inference. * Compressed video: The model operates on 32-channel DCAE-V latents, a compact representation of the source frames. * Video decoding: A fine-tuned PixelUMM decoder converts those latents back into frames. * Prompt processing: A separate Nomas-PE model turns raw instructions into structured generation prompts during interactive sessions. * Domain adapters: The repository includes a cinematic LoRA and a robotics LoRA trained on the RealSource-World dual-arm dataset. LoRA adapters modify a model through small trainable matrices, reducing the cost of domain adaptation. Three checkpoints target different jobs. | Checkpoint | Precision | Primary use | | Base DiT | FP32, non-distilled | LoRA fine-tuning | | Distilled DiT | FP32 | Four-step, six-step, or eight-step inference | | Quantized distilled DiT | INT8 | Low-memory inference on an RTX 4090 | How the RTX 4090 configuration saves memory. The dedicated low-memory configuration combines quantization, bounded history, and staged CPU offloading to fit inference on one RTX 4090. Its main settings are: * INT8 storage: DiT weights and the KV cache use INT8, while linear and attention calculations remain in BF16. * Sliding history: A fixed context window bounds KV-cache growth. This reduces memory use by limiting how much earlier history remains directly available. * CPU offloading: The text encoder, DiT, and VAE move between system memory and GPU memory as each stage runs. * Optimized kernels: The configuration uses FlexAttention for the DiT and CUTLASS FP8 for video decoding. * No sequence parallelism: The single-GPU path disables the sequence splitting used by the multi-GPU configuration. Default multi-GPU inference enables sequence parallelism, which distributes sequence processing across devices. The supplied supervised fine-tuning configuration uses eight-way sequence parallelism. On 128 GPUs, that produces 16 data-parallel replicas, each spanning eight GPUs. The reference LoRA recipe sets both rank and alpha to 256 and trains for 2,000 steps with the Muon optimizer. Fine-tuning keeps the fast inference path. Hy-Nomas separates adaptation from deployment. Developers train a LoRA against the non-distilled base checkpoint, then merge the resulting adapter into the distilled checkpoint used for inference. This workflow preserves access to the distilled model's four-step, six-step, and eight-step generation modes. * Caption each video segment and attach any normalized camera trajectories. * Encode the frames into DCAE-V latents. * Package the processed samples as Arrow or Parquet data for supervised fine-tuning. * Train the LoRA against the base checkpoint. * Merge the adapter into the distilled checkpoint for deployment. From packed data to merged weights. python scripts/data_processing/build_sft_packing_plan.py \ -input /path/to/train.arrow \ -output /path/to/train_packing.json \ -max-seq-len 360000 \ -num-groups 16 bash scripts/train/train_from_config.sh configs/experiment/sft.yaml \ dit_weight="$Hy_Nomas_Model_Path/DiT/base_ar/...safetensors" \ dataset.data_file=/path/to/train.arrow python -m scripts.train.merge_lora \ -base "$Hy_Nomas_Model_Path/DiT/distilled/...safetensors" \ -adapter ./outputs/segment_lora/adapter \ -output ./outputs/merged_model.safetensors Where developers can apply it. * Robotics research: The dual-arm adapter can support synthetic observation generation and controlled scenario exploration. Generated video still requires validation before use as physically accurate simulation or policy-training data. * Interactive prototypes: Text and camera controls can drive game-like scenes without requiring a fixed clip to be regenerated from the beginning. * Cinematic planning: The cinematic adapter supports iterative previs and storyboard development. * Long-horizon video research: The exposed KV-cache implementation gives researchers a concrete system for studying memory, drift, and temporal consistency. The benchmarks have not arrived. At publication, the GitHub repository had a single commit and listed the technical report as forthcoming. No public benchmark results establish prompt adherence, camera-control accuracy, generation latency, physical consistency, or long-horizon stability against systems such as Genie 3, Wan, or Veo. The low-memory path also leaves practical deployment questions unanswered. CPU offloading requires additional system RAM and introduces CPU-GPU transfer overhead, but the release does not specify the required RAM or measured throughput on an RTX 4090. The quality impact of INT8 weights, an INT8 KV cache, and a sliding history window also remains unquantified. Hy-Nomas gives developers an inspectable stack for interactive video generation, including model weights, data preparation, training, LoRA adaptation, and local inference. The Apache 2.0 license also permits modification and commercial use under its terms. Developers can test the interactive demo or download the code from the GitHub repository.