The original title is: "FFmpeg 8.0 in practice: transcribe with the Whisper filter and probe av1_vulkan"

Short summary

FFmpeg 8.0 adds a native Whisper audio filter for in-pipeline speech-to-text and av1_vulkan for cross-vendor GPU AV1 encoding. The post demonstrates generating SRT/JSON captions in a single FFmpeg command and building a script that probes for av1_vulkan with automatic fallback to SVT-AV1. These features eliminate the need for separate transcription services and vendor-specific GPU encoders.

• •FFmpeg 8.0 adds native Whisper transcription as an audio filter, replacing separate STT services
• •av1_vulkan enables cross-vendor GPU AV1 encoding without NVENC/QSV lock-in
• •Includes queue tuning guidance and automatic fallback to CPU SVT-AV1

Generated with AI, which can make mistakes.