| --- |
| title: Turntable Time Machine |
| emoji: 📻 |
| colorFrom: yellow |
| colorTo: purple |
| sdk: gradio |
| sdk_version: 5.49.1 |
| python_version: 3.10.13 |
| app_file: app.py |
| pinned: false |
| license: mit |
| models: |
| - ACE-Step/Ace-Step1.5 |
| - facebook/mms-tts-eng |
| - facebook/mms-tts-deu |
| - facebook/mms-tts-fra |
| - facebook/mms-tts-spa |
| - CohereLabs/tiny-aya-global |
| tags: |
| - build-small |
| - thousand-token-wood |
| - gradio |
| - small-models |
| - multilingual |
| - audio |
| - music-generation |
| - text-to-audio |
| - track:wood |
| - sponsor:openai |
| - achievement:offgrid |
| - achievement:offbrand |
| - achievement:fieldnotes |
| --- |
| |
| # 🤗 💿📻⏳ Turntable Time Machine |
|
|
| ## **Pick two eras 🕰️. Summon a fictional DJ 🎙️. Hear a tiny radio broadcast from a timeline that never happened 📻✨.** |
|
|
| Turntable Time Machine is a whimsical Gradio app built for the Hugging Face x Gradio Build Small Hackathon. It lets users choose a source era 🕰️, remix era 🔁, fictional DJ host 🎙️, language 🌍, mood 🌈, texture 📼, and vocal mode 🎵, then generates a short original AI radio postcard 📻. |
|
|
| *The twist:* this is not a famous-song remix machine. The app uses broad musical-era vocabularies and safety constraints to make new little audio artifacts: 1960s soul through a 1990s house lens, 1980s synth-pop polished into 2000s filtered disco house, or a 2020s lo-fi signal drifting backward through old radio static. |
|
|
| This submission targets the **Thousand Token Wood** track: playful, compact, AI-native, and built around small/open model targets. ACE-Step, MMS TTS, and Tiny Aya are used as optional model hooks, while deterministic templates and fallback audio keep the app demonstrable when model access or GPU runtime is limited. |
|
|
| ## 🔗 Submission Links |
|
|
| - 🤗 Live Space: https://huggingface.co/spaces/build-small-hackathon/turntable-time-machine |
| - 💻 GitHub repo: https://github.com/jd-coderepos/turntable-time-machine |
| - 🎬 Demo video: https://www.youtube.com/watch?v=EKCwTK88pUE |
| - 📣Social Post on LinkedIn: [Trying to spread the word about the app](https://www.linkedin.com/posts/jennifer-l-dsouza_buildsmallhackathon-huggingface-gradio-share-7472410854131527680-2uYV/?utm_source=share&utm_medium=member_desktop&rcm=ACoAAAJH7gwBPr4GzY6aqpAOF4R6kX-N1rYsgZ4) 🤗 |
| - 📝Field notes: [Bending Time with a Virtual DJ -- Field Notes from the Build Small AI Hackathon](https://medium.com/@jenlindadsouza/bending-musical-time-with-a-fictional-dj-field-notes-from-the-build-small-hackathon-71f10680ca0e?postPublishedType=repub) |
|
|
| ## 🎬 Demo Script |
|
|
| A concise demo flow: |
|
|
| 1. 🤗 Open the [Turntable Time Machine Space](https://huggingface.co/spaces/build-small-hackathon/turntable-time-machine). |
| 2. Set **Source era** to **1960s** and **Source genre** to **Folk pop**. |
| 3. Set **Remix era** to **2000s** and **Remix genre** to **2000s indie rock**. |
| 4. Choose **The Warehouse Announcer** as the fictional DJ persona. |
| 5. Choose **French** as the broadcast language. |
| 6. Choose **Original micro-lyrics** and keep **Night drive** as the lyric theme. |
| 7. Set **Music prompt language** to **English + broadcast language** to show the app's bilingual prompt path. |
| 8. Keep **nostalgic and warm**, choose **Clean digital master**, and use the **10 second** duration. |
| 9. Click **Bend the Timeline**. |
| 10. Play the **Final broadcast**, then compare it with **Music only**. |
| 11. Show the **Fictional DJ intro** text card and the **Spoken DJ intro** audio when MMS TTS is available. |
| 12. Show the **Generated micro-lyrics** card and the **Mixtape Card** with the route, model status, BPM, seed, texture, and safety note. |
|
|
| 🎬 Recommended one-video case: **1960s Folk pop -> 2000s indie rock** with **The Warehouse Announcer**, **French**, **Original micro-lyrics**, and **English + broadcast language**. It shows the app's full loop: era routing, fictional host text, multilingual prompt routing, generated micro-lyrics, safety-aware prompt, texture processing, and final audio. |
|
|
| ⚠️ Note: The app does **not** generate covers, real-DJ imitation, voice clones, copyrighted lyrics, or famous melodies. Outputs are original compositions guided by broad genre descriptors. |
|
|
| ## ℹ️ Main Features |
|
|
| * Custom illustrated [Gradio](https://www.gradio.app/) interface |
| * Era-to-era routing across the 1960s, 1970s, 1980s, 1990s, 2000s, 2010s, and 2020s |
| * Curated genre vocabulary for soul, disco, funk, synth-pop, house, trip-hop, UK garage, hyperpop, lo-fi, amapiano-inspired grooves, and AI-era electronic |
| * Six fictional DJ personas with original intro templates |
| * English, German, French, and Spanish broadcast text support |
| * Three vocal modes: Instrumental only, Wordless vocal texture, and Original micro-lyrics |
| * Nostalgia textures: vinyl crackle, cassette warmth, FM radio, mono radio, club PA, VHS glow, early MP3 compression, and clean digital master |
| * Surprise Me route generation |
| * Mixtape card with route, language, texture, BPM, seed, model status, and safety note |
| * Lazy model hooks and fallback generation for deployment reliability |
|
|
| ### 🧠 Built-in Timeline Engine |
|
|
| Turntable Time Machine is not just a prompt wrapper around an audio model. Before generation, it runs a lightweight programmed Timeline Engine: |
|
|
| ```text |
| User selections |
| → source era / genre lookup |
| → remix era / genre lookup |
| → fictional DJ persona selection |
| → language and prompt-mode routing |
| → BPM selection from genre ranges |
| → optional Tiny Aya micro-lyrics |
| → safety-aware music prompt |
| → optional MMS TTS DJ narration |
| → ACE-Step music hook or fallback synth audio |
| → nostalgia texture processing |
| → final broadcast and mixtape card |
| ``` |
|
|
| 🛠️ The design explores a simple idea: small models work better when the app gives them a playful structure. Curated taxonomies, deterministic routing, and clear guardrails turn a broad generation task into a tiny musical instrument. |
|
|
| ## 🤖 Models Used |
|
|
| | Component | Model | Parameters | Used for | |
| | --- | --- | ---: | --- | |
| | Music generation target | [ACE-Step/Ace-Step1.5](https://huggingface.co/ACE-Step/Ace-Step1.5) | under 32B target stack | Original short music clips | |
| | Spoken DJ intro target | [facebook/mms-tts-eng](https://huggingface.co/facebook/mms-tts-eng), `deu`, `fra`, `spa` variants | 36.3M each for English model card | Optional fictional DJ narration in the selected broadcast language | |
| | Text / micro-lyrics target | [CohereLabs/tiny-aya-global](https://huggingface.co/CohereLabs/tiny-aya-global) | small multilingual model | Optional two-line original lyrics | |
| | Template and synth fallback | none | 0 | Keeps the app usable in restricted runtimes | |
|
|
| ACE-Step and MMS TTS are lazy optional hooks. If model packages, weights, or hardware are unavailable, the app falls back to playable synthetic demo audio and text-only DJ intros. Tiny Aya falls back to deterministic micro-lyrics when access is restricted. |
|
|
| ## 🗂️ Data |
|
|
| The app's character layer lives in `data/`: |
|
|
| * `music_eras.json` defines eras, genres, safe prompt styles, BPM ranges, instruments, sonic traits, and avoid notes. |
| * `dj_personas.json` defines six fictional hosts, including The Midnight Radio Curator and The Neon Signal Host. |
| * `audio_textures.json` defines record, cassette, radio, club, VHS, MP3, and clean-master post-processing flavors. |
| * `languages.json` defines English, German, French, and Spanish. |
| * `lyric_themes.json` defines compact lyric themes such as Future city, Night drive, Radio signal, and Time travel. |
|
|
| ## 🛠️ Local Setup |
|
|
| ```bash |
| python -m venv .venv |
| source .venv/bin/activate # Windows: .venv\Scripts\activate |
| pip install -r requirements.txt |
| python app.py |
| ``` |
|
|
| Run tests: |
|
|
| ```bash |
| python -m unittest discover -s tests |
| ``` |
|
|
| For mocked model paths in tests only: |
|
|
| ```bash |
| set TTM_TEST_MODEL_CALLS=1 |
| python -m unittest discover -s tests |
| ``` |
|
|
| ## Hugging Face Space / ZeroGPU Notes |
|
|
| Recommended Space settings: |
|
|
| ```text |
| SDK: Gradio |
| Python: 3.10.13 |
| App file: app.py |
| Hardware: ZeroGPU or compatible GPU runtime |
| ``` |
|
|
| Useful environment variables: |
|
|
| ```text |
| TEXT_MODEL_ID=CohereLabs/tiny-aya-global |
| HF_TOKEN=hf_your_token_here |
| HUGGING_FACE_HUB_TOKEN=hf_your_token_here |
| ``` |
|
|
| If model access fails, add an `HF_TOKEN` secret with read access and make sure the Space owner has accepted any required model terms. |
|
|
| ## 🛡️ Protective Guardrails |
|
|
| Turntable Time Machine is designed for original audio from broad musical-era descriptors. Its prompt builder avoids requests to imitate specific songs, artists, singers, producers, DJs, hooks, samples, lyrics, melodies, or arrangements. |
|
|
| The lyric generator asks for only two short original lines and sanitizes generated text before use. The DJ personas are fictional hosts, not imitations of real broadcasters or performers. |
|
|
| ## 🙌 Acknowledgements |
|
|
| This project was built using [**OpenAI Codex**](https://openai.com/codex/) as a coding agent. Codex supported the Gradio app structure, timeline orchestration logic, fallback audio behavior, safety checks, tests, and submission-readiness polish. |
|
|
| Thank you to the **Gradio team** and the [Build Small Hackathon](https://huggingface.co/build-small-hackathon) partners for making room for tiny, playful, local/open-weight AI experiments. 🎛️📻✨ |
|
|
