How Tencent’s TXAV1 Encoder Achieves 20% Bandwidth Savings and Industry‑Leading Quality
This article details Tencent Cloud's self‑developed TXAV1 AV1 encoder, covering its evolution from V265, performance gains over competing codecs, compression‑rate improvements through pre‑analysis models, engineering acceleration techniques, and real‑world deployment in live streaming, video‑on‑demand, and AVIF image services.
1. TXAV1 Codec Capabilities and Applications
Since 2017 Tencent has been developing its own codecs, culminating in the TXAV1 AV1 encoder. Building on V265 optimizations, TXAV1 is deployed across Tencent Cloud, serving domestic and overseas customers, supporting 60 fps live streams and 8K30 fps video, and achieving notable bandwidth savings.
TXAV1 demonstrates superior compression compared to SVT‑AV1, AOM‑3.2, and VVenc, delivering 12‑20% bandwidth reduction while maintaining or improving visual quality. Subjective tests show TXAV1 outperforms H.265 at similar bitrates, and it consistently ranks first in AV1 benchmark competitions.
Beyond video, TXAV1 supports AVIF image encoding, offering over 30% bandwidth savings versus WebP and significant improvements over HEIF, with optimized memory handling for high‑resolution images.
2. New Pre‑analysis Model for Compression Improvement
The encoder employs a three‑pronged approach: enhanced reference‑frame management, full‑dimensional adaptive quantization, and rate‑distortion model optimization. By analyzing frame dependencies and block complexities, TXAV1 achieves up to 2% additional compression gain while enabling parallel processing of the TPL (temporal prediction loop) stage.
Parallelizing TPL without reconstruction reduces CPU usage by up to 50% in fast‑encode modes, though it sacrifices some reconstruction‑based accuracy.
3. TXAV1 Engineering Acceleration Techniques
Data‑structure redesign addresses the overhead of deeper CU partitioning, cache misses, redundant encoding, and buffer copying. Strategies include fixed‑type CTU addressing, local buffers for neighboring CU data, identical‑CU reuse, and swap buffers to avoid copies.
Parallelism is realized through five modes: pre‑analysis, inter‑frame, WPP (wavefront parallel processing), post‑processing, and entropy coding. TXAV1 achieves up to 334% speedup with only 0.28% compression loss, outperforming SVT‑AV1 and AOM.
Various WPP configurations (no‑WPP, full‑fake, half‑fake, half‑real, full‑real) are employed adaptively, balancing speed and quality.
4. Summary
TXAV1 now offers industry‑leading AV1 video and image coding capabilities, integrated into Tencent Cloud MPS products, delivering 20%+ bandwidth savings, high visual fidelity, and efficient deployment across cloud services.
Tencent Architect
We share technical insights on storage, computing, and access, and explore industry-leading product technologies together.
How this landed with the community
Was this worth your time?
0 Comments
Thoughtful readers leave field notes, pushback, and hard-won operational detail here.