LG Electronics

The world's first AI-powered 24/7 song dedication station

Consumer Electronics

Entertainment

Advertising & Marketing

Client
LG Electronics
Expertise

AI Development

Full-Stack Engineering

Multi-Modal AI Orchestration

Tech Stack

OpenAI

Suno.AI

AWS Lambda

Challenge

A Brand Promise That Demanded Real Technology

LG Electronics runs one of the most ambitious brand platforms in consumer electronics. Built around the tagline "Life's Good," the company's marketing campaigns go beyond product promotion to address real social issues. In 2025, LG commissioned new research on human connection and the findings were stark: 68% of people find it harder to make real friends than in the past, and a third reported having one or fewer meaningful connections in the previous month.

Working with TBWA\Chiat\Day New York (part of Omnicom), LG conceived Radio Optimism — a 24/7 AI-powered radio station where anyone in the world could send a personalized song dedication to someone they care about. The creative concept was clear. The technical challenge was anything but.

The Engineering Scale Problem

The brief called for something that had never been built before: a consumer-facing product that chains together three distinct AI generation steps — lyrics, music, and visual art — into a single user experience that feels immediate and personal. Each dedication needed to produce a genuinely unique output. Not a template with name substitution. Not a remix of existing tracks. A fully original song with custom lyrics drawn from the user's personal message, composed in a genre of their choosing, paired with artwork that reflects the song's emotional tone.

The scale requirements compounded the complexity. This was a global campaign for one of the world's largest electronics brands. The system needed to handle thousands of concurrent users, support multiple languages from launch, and maintain consistent quality across every single generation.

Multi-Modal AI Orchestration at Production Quality

At the time of development, most multi-modal AI applications existed as demos or internal tools. Deploying three AI services — text generation, audio composition, and image creation — in a chained production pipeline serving consumer traffic was uncharted territory. Each model has its own latency profile, failure modes, and quality variance. Orchestrating them into something that feels like a single, smooth user experience required a purpose-built infrastructure layer.

Solution

From Creative Brief to Technical Architecture

Twistag was brought in as the engineering partner responsible for the full technical build. TBWA\Chiat\Day New York owned the creative concept, brand strategy, and campaign execution. Twistag's mandate: turn the creative vision into a production-ready AI platform that could serve a global audience from day one.

The project started with a joint architecture workshop where Twistag mapped the user journey against the AI capabilities available. The core design decision was to treat the system as a pipeline rather than a monolith — each AI service operates independently, connected by an orchestration layer that manages sequencing, error handling, and quality control.

Multi-Modal AI Pipeline Architecture

The core of Radio Optimism is a multi-modal AI pipeline that transforms a few sentences of personal text into three distinct creative outputs: lyrics, a fully composed song, and matching album art.

Step 1: Dedication Intake and Processing — The user provides three inputs: the name of the person they are dedicating the song to, a short personal message, and a genre/vibe selection. The intake service validates the input, detects the language (supporting six languages at launch), and structures the data for downstream AI services.

Step 2: AI Lyric Generation — The structured input passes to OpenAI's language model with a carefully engineered prompt chain. The prompt system extracts the emotional core of the user's message, maps it to song structure conventions for the selected genre, and generates lyrics with proper verse/chorus/bridge formatting. Automated quality scoring evaluates each generation for structural completeness and emotional alignment.

Step 3: AI Music Composition — The generated lyrics, along with genre and mood metadata, pass to Suno.AI for full audio composition. Twistag's orchestration layer manages the handoff between lyric generation and music composition, using asynchronous processing with webhook callbacks to avoid blocking.

Step 4: AI Album Art Creation — In parallel with music composition, Adobe Firefly generates album art based on the song's emotional profile, genre, and thematic elements. Running this step in parallel reduces total pipeline time significantly.

Step 5: Assembly and Publishing — Once all three AI outputs are ready, the assembly service combines them into a single playable package with a unique shareable URL, publishing to the 24/7 Radio Optimism streaming playlist.

Global Infrastructure on AWS Lambda

The entire pipeline runs on AWS Lambda, giving the system automatic scaling to match demand and zero cost when idle. Twistag designed the Lambda functions to be stateless, with all generation state managed through a coordination service — no session affinity, no bottlenecks.

Multi-Language Support

Supporting six languages was not just a translation task. Twistag built language-specific prompt configurations for the lyric generation stage, tuning each one for natural-sounding output in that language rather than generating in English and translating.

Quality and Safety

For a global consumer brand, every generated song represents LG. Twistag implemented a multi-layer quality system: automated scoring at the lyric generation stage (structural and emotional alignment), output validation at the music composition stage (audio completeness and genre accuracy), and content safety filtering across all text and image outputs. The system catches and retries poor-quality generations before they reach the user, without adding perceptible latency.

No items found.

Results

Key Results:

  • 700,000+ unique, personalized songs created
  • 2.4 billion content views globally
  • 30 million social engagements worldwide
  • 1.5 million listens and shares
  • 6 languages supported at launch
  • 2:1 share-to-creation ratio — the strongest signal of genuine user value

The platform created over 700,000 completely unique songs — each one with original lyrics derived from a personal dedication, a fully composed audio track in the user's selected genre, and custom album art. This is not 700,000 variations of the same template. Each song is a distinct creative output.

The campaign's 2.4 billion content views and 30 million social engagements indicate that the generated songs were not just created but actively shared, listened to, and talked about. The 1.5 million listens and shares against 700,000 creations produces a share-to-creation ratio above 2:1.

Operating across six languages from a single platform proved that generative AI consumer products can scale globally without multiplicative engineering effort.

The campaign attracted coverage from Ad Age, Marketing Dive, LBBOnline, and dozens of other industry publications. Radio Optimism demonstrated that generative AI can power emotionally resonant brand experiences — not just efficiency tools.

Let's meet.

Whether you need a dedicated team, end-to-end project, or a proof of concept, we're flexible to support your journey.

Book a Discovery Call

You’ll be speaking with one
of our tech project managers.