#ai

Posts on X2Q tagged ai. Updated as new posts are published.

• Jun 7, 2026

  Extending video clips with AI on a 12 GB GPU — six models compared

  You have a short machine/action clip and want it longer. 'Extending' a clip is really image-to-video: take the last frame and generate a plausible continuation. I fed the same shot to six models — Wan2.2-TI2V-5B, Wan2.2-I2V-A14B (GGUF), Stable Video Diffusion, LTX-Video and CogVideoX-5B — all on a single 12 GB RTX 4070 Ti. The interesting part isn't which looks best; it's the memory-offload tricks that make a 14B video model (and 720p) fit in 12 GB at all.

• Jun 4, 2026

  Local speech-to-text with whisper.cpp — a self-hosted Google Speech API alternative (2026)

  Years ago I ran transcription through the Google Speech API. In 2026 you don't need a cloud API for that — whisper.cpp runs OpenAI's Whisper models locally on a laptop, CPU or GPU, with no keys and nothing leaving the machine. Here's the full path from clone to transcript.

• Jun 2, 2026

  Five ways to get b-roll from one ride — all on a local GPU

  B-roll is the cutaway footage that makes an edit breathe — and onboard POV footage from a single camera has none. I tried five ways to manufacture it from one ATV ride: CV auto-mining of the real footage, generative stills (SDXL-Turbo → RealVisXL img2img), AI Ken Burns via Depth-Anything-V2 parallax, true generated motion via Stable Video Diffusion, and optical-flow slow-motion. All local on a 12 GB card. Here's what each is good for, and the gotchas.