Nano Banana → Veo → Lyria Pipeline: Multimodal Generative Workflow Guide

PromptCrates tests and documents end-to-end production pipelines across state-of-the-art models. While single-model prompting produces impressive one-off outputs, commercial workflows require coordinating specialized tools: visual keyframing, temporal motion synthesis, and adaptive musical accompaniment.

The Three-Stage Production Architecture

In modern generative media, three distinct foundation models handle these specialized responsibilities:

• 1. Nano Banana (Image Layer) — High-fidelity text-to-image and character consistency engine used to establish precise visual composition, wardrobe, lighting, and camera angle as a locked reference keyframe.
• 2. Google Veo (Video Layer) — Cinematic motion model that ingests the Nano Banana keyframe to simulate continuous physical dynamics, realistic camera pans, fluid simulations, and temporal consistency.
• 3. Google Lyria (Audio Layer) — Advanced music and acoustic synthesis model that composes synchronized instrumental beds, thematic motifs, and ambient textures matching the pacing and emotional tone of the video.

Stage 1: Precision Keyframing in Nano Banana

Every coherent video sequence begins with an anchor image. Using the PromptCrates three-layer structure (Subject · Atmosphere · Technical), prompt Nano Banana to output a high-resolution 16:9 cinematic still.

1. Establish the Subject: Clearly specify character anatomy, costume materials, facial expression, and spatial orientation. Lock physical descriptors to prevent drift across shots.
2. Calibrate the Atmosphere: Set directional lighting, atmospheric haze, color palette (e.g., volumetric golden hour or neon cyber-noir), and spatial depth.
3. Lock Technical Flags: Enforce 16:9 widescreen aspect ratio, cinematic lens simulation (e.g., 35mm anamorphic, shallow depth of field), and neutral focal plane.

Stage 2: Motion Animation & Camera Direction in Google Veo

Import the locked Nano Banana image as the first-frame conditioning anchor into Google Veo. Rather than re-describing the visual appearance, focus your Veo prompt purely on dynamic velocity, camera trajectory, and physical movement.

• Camera Choreography: Direct the lens explicitly (e.g., slow push-in, subtle orbital pan, low-angle tracking shot).
• Environmental Motion: Describe secondary motion such as wind rustling foliage, particulate drift, rain reflections, or fabric ripple.
• Temporal Physics: Keep acceleration curves natural and specify temporal duration (e.g., 5–8 second seamless progression).

Stage 3: Soundtrack & Acoustic Scoring in Google Lyria

Once the video timeline is locked, generate an original soundtrack using Google Lyria. Map the acoustic prompt using the PromptCrates music framework (Theme · Mood · Production).

• Theme: Select acoustic instrumentation and melodic motif (e.g., minimalist cinematic piano, pulsing analog synthesizer, or orchestral strings).
• Mood: Match the emotional trajectory of the Veo video clip (e.g., contemplative, rising tension, or expansive wonder).
• Production: Lock tempo (BPM), mix space (reverb decay, stereo width), and arrangement dynamics to hit visual cut points.

Connecting the Pipeline in Post-Production

Combine the generated assets in your editor or automated pipeline. Because the keyframe established the visual fidelity and lighting, the Veo video inherits consistent color grading, allowing the Lyria audio bed to snap into rhythm with zero visual discord. Explore our companion directories and frameworks in the AI Tools Directory and Citation Guide.