AI Summary
→ WHAT IT COVERS Host Claire documents a live experiment using Google Flow and the Gemini Omni video model to build a one-minute AI avatar hype video for her podcast. Starting with zero tool knowledge, she completes the full workflow — avatar creation, storyboard generation, video rendering, and timeline editing — in under fifteen minutes. → KEY INSIGHTS - **Avatar capture speed:** Google Flow's mobile QR code scanning process captures a usable facial avatar in under two minutes, requiring only frontal and side-profile head turns. The system automatically pulls background details from the scan environment — posters, books, wall color — and incorporates them into generated scenes without additional prompting. - **AI as creative director:** Rather than jumping straight to video generation, prompting Flow to build a storyboard first produces a structured seven-scene shot list with specific camera directions, lighting notes, and character blocking. This intermediate step prevents generic output and gives non-video-literate creators a professional production framework before a single frame renders. - **Dual-version rendering:** Flow automatically generates two versions of every video clip simultaneously, mirroring Veo 2 behavior. Reviewing both versions per scene and selecting the stronger take before editing meaningfully improves final output quality without additional generation cost or time investment. - **Character consistency limitations:** At current capability, the avatar matches the source face roughly 50% of the time across scenes. Hair length, background color, shelf contents, and lighting shift between clips. Mitigation strategy: use consistent background descriptors in every scene prompt and supply multiple reference images to the Omni model to tighten character coherence. - **Browser-native timeline editing:** Flow includes a built-in video editor accessible directly in the browser, eliminating the need for external software. Stitching seven AI-generated scenes into a finished one-minute video takes approximately five minutes by dragging clips into the storyboard-specified sequence and selecting preferred takes per scene. → NOTABLE MOMENT When the avatar video rendered, it accurately reproduced a specific NVIDIA product visible only in the background of Claire's avatar scan photos — a detail she had not mentioned in any prompt. The model extracted and placed environmental context from the original capture without instruction. 💼 SPONSORS [{"name": "Merge", "url": "https://merge.dev/howiai"}, {"name": "Jira Product Discovery", "url": "https://atlassian.com/howiai"}] 🏷️ AI Video Generation, Google Flow, Gemini Omni, AI Avatars, Generative Media Tools