AuraLuna

• 

• 

  Logo

• 

Inspiration

ASMR is primarily known for its role in relaxation, stress relief, and sleep induction. However, a growing body of research suggests that the unique state induced by ASMR may have measurable effects on cognitive function, including attention, focus, and memory.

What it does

AuraLuna creates on-demand ASMR for calm in stressful situations as well as sleep, utilizing AI - Claude for creative text generation and Fish Audio for cloned voice synthesis.

There are several drop downs available which give generalized ASMR themes, but one specifically is called 'Calm' which was created to help with public anxiety by exuding calm and reminding the user to breathe. This was done with the intention of helping the user (who might be alone at that moment) work through panic.

How I built it

AuraLuna was built as a Next.js web application, utilizing a hybrid architecture combining an interactive frontend with server-less backend API routes to manage two different AI providers: Claude (for creative prompts) and Fish Audio (for static TTS).

User Experience

Logo: Headphones with sound waves worn by a crescent moon, reclining on a cloud. Palette: Soothing and serene, soft blues, purples, muted grays. User Input: The user selects a Mood (e.g., "Ocean," "Sleepy", "Calm") and clicks "Generate Whisper." Content Creation: When the client sends a request with a specific mood (e.g., "Calm"), the backend sends a prompt to Claude (Anthropic). Claude is instructed to act as a sleep or anxiety relief narrator and create a short soothing script based on the user's chosen mood. The cloned voice reads the script aloud in a measured tone.

Challenges I ran into

None of the voices on Fish.Audio were soft and slow enough for my purposes. I finally was able to train the model on one of my favorite ASMR artists, ALB. As such, the audio content generated by AuraLuna's text-to-speech (TTS) services is provided for personal, non-commercial use only. Please see additional information in the README under "Voice Generation Use Policy".

Additionally, I had difficulty getting cloned voice to be used in the app instead of the rotation of stock voices available. This was solved after much wheel spinning by using reference_id, alongside model_id.

Accomplishments that I am proud of

I was a solo team who only started the CalHacks Hackathon work on Saturday and thus relied on Gemini and OpenAi to help round out my knowledge gaps. I loved the logo I was able to design through using a mix of Gemini, OpenAi, and Canva and was able to achieve a soothing and simple color palette through the use of Tailwind and color theory research.

Built With

• anthropic
• canva
• gemini
• javascript
• netlify
• nextjs
• openai
• tailwind