Build a song generation app using audio.cpp, supporting low VRAM machines
@cocktailpeanutBuild me a beautiful, fully functional, locally running AI song generation web app powered by audio.cpp, with a cross-platform Pinokio launcher.
First, interview me before writing any code.
Ask me a few simple questions to understand:
Keep the interview friendly, concise, and interactive. Ask follow-up questions when necessary, but don't overwhelm me with technical details. Help me make decisions rather than expecting me to know everything.
Once you understand what I want, build it.
Follow existing Pinokio launcher conventions and audio.cpp's official documentation. Prioritize a working, delightful experience over unnecessary complexity.
Finally, offer to publish the finished app.
Once the app is working, ask me if I'd like to publish it to GitHub so others can install and run it through Pinokio.
If I say yes:
Include an install badge in the README.md.
Add the following badge near the top of README.md, replacing OWNER/REPO with the actual GitHub repository:
[](https://pinokio.co/resolve?url=https://github.com/OWNER/REPO)
After publishing the GitHub repository, give the user the https://pinokio.co/resolve?url=https://github.com/OWNER/REPO install link and ask them to open it to add the project if prompted, signing in if needed.
Don't publish anything without my explicit approval.
Start by interviewing me. Don't write code until we've established what we're building.
Your selected agent may use an external provider. Execution uses that agent’s permissions.
Fetch this JSON template and carry out its instructions: https://pinokio.co/templates/01m76khmj5dzjffechznxdcr59/definition.json
The agent receives the template link and your answers, then reads and applies the template. Its interpretation can differ from this preview.
Published instructions cannot be edited. Linked resources can still change. Review the instructions before running an agent.
Build me a beautiful, fully functional, locally running AI song generation web app powered by [audio.cpp](https://github.com/0xShug0/audio.cpp), with a cross-platform Pinokio launcher. **First, interview me before writing any code.** Ask me a few simple questions to understand: 1. **What I want to build:** What kind of music generation app do I envision? What should users be able to do? Suggest useful features and help me brainstorm if I'm unsure. 2. **Which AI model to use:** Research the music generation models currently supported by the latest version of audio.cpp (such as ACE-Step, YuE2, HeartMuLa, MiniMax Music, and others). Recommend suitable options, briefly explaining their capabilities, limitations, and hardware requirements. Help me choose. 3. **What it should look like:** Ask about my preferred visual style, layout, and overall experience. Invite me to share inspiration through screenshots, images, URLs, or written descriptions. If I don't have a design in mind, suggest a few creative directions. Keep the interview friendly, concise, and interactive. Ask follow-up questions when necessary, but don't overwhelm me with technical details. Help me make decisions rather than expecting me to know everything. **Once you understand what I want, build it.** - Create a polished, responsive web UI based on my preferences. - Integrate the selected model through audio.cpp's supported interfaces. - Build a Pinokio launcher that handles installation, model downloads, and starting the app with one click. - Support Windows, macOS, and Linux, automatically detecting available hardware acceleration and choosing appropriate backends. - Make the app fully local, without requiring cloud APIs. - Include essential features such as generation progress, audio playback, and downloading generated songs. - Verify that everything works, handle errors gracefully, and don't expose features unsupported by the selected model. Follow existing Pinokio launcher conventions and audio.cpp's official documentation. Prioritize a working, delightful experience over unnecessary complexity. **Finally, offer to publish the finished app.** Once the app is working, ask me if I'd like to publish it to GitHub so others can install and run it through Pinokio. If I say yes: 1. Help me choose a repository name and description. 2. Create a GitHub repository under my account, with my approval. 3. Include the complete source code, Pinokio launcher, README, installation instructions, and appropriate license. 4. Make sure the repository is ready for others to clone, install, and launch through Pinokio. 5. Commit and push everything to GitHub, after confirming the repository's visibility and contents with me. 6. Give me the published repository URL. **Include an install badge in the README.md.** Add the following badge near the top of README.md, replacing OWNER/REPO with the actual GitHub repository: ``` [](https://pinokio.co/resolve?url=https://github.com/OWNER/REPO) ``` After publishing the GitHub repository, give the user the https://pinokio.co/resolve?url=https://github.com/OWNER/REPO install link and ask them to open it to add the project if prompted, signing in if needed. **Don't publish anything without my explicit approval.** **Start by interviewing me. Don't write code until we've established what we're building.**
{
"title": "Build a cross platform, low VRAM song generation app",
"description": "Build a song generation app using audio.cpp, supporting low VRAM machines",
"instructions": "Build me a beautiful, fully functional, locally running AI song generation web app powered by [audio.cpp](https://github.com/0xShug0/audio.cpp), with a cross-platform Pinokio launcher.\n\n**First, interview me before writing any code.**\n\nAsk me a few simple questions to understand:\n\n1. **What I want to build:** What kind of music generation app do I envision? What should users be able to do? Suggest useful features and help me brainstorm if I'm unsure.\n2. **Which AI model to use:** Research the music generation models currently supported by the latest version of audio.cpp (such as ACE-Step, YuE2, HeartMuLa, MiniMax Music, and others). Recommend suitable options, briefly explaining their capabilities, limitations, and hardware requirements. Help me choose.\n3. **What it should look like:** Ask about my preferred visual style, layout, and overall experience. Invite me to share inspiration through screenshots, images, URLs, or written descriptions. If I don't have a design in mind, suggest a few creative directions.\n\nKeep the interview friendly, concise, and interactive. Ask follow-up questions when necessary, but don't overwhelm me with technical details. Help me make decisions rather than expecting me to know everything.\n\n**Once you understand what I want, build it.**\n\n- Create a polished, responsive web UI based on my preferences.\n- Integrate the selected model through audio.cpp's supported interfaces.\n- Build a Pinokio launcher that handles installation, model downloads, and starting the app with one click.\n- Support Windows, macOS, and Linux, automatically detecting available hardware acceleration and choosing appropriate backends.\n- Make the app fully local, without requiring cloud APIs.\n- Include essential features such as generation progress, audio playback, and downloading generated songs.\n- Verify that everything works, handle errors gracefully, and don't expose features unsupported by the selected model.\n\nFollow existing Pinokio launcher conventions and audio.cpp's official documentation. Prioritize a working, delightful experience over unnecessary complexity.\n\n\n**Finally, offer to publish the finished app.**\n\nOnce the app is working, ask me if I'd like to publish it to GitHub so others can install and run it through Pinokio.\n\nIf I say yes:\n\n1. Help me choose a repository name and description.\n2. Create a GitHub repository under my account, with my approval.\n3. Include the complete source code, Pinokio launcher, README, installation instructions, and appropriate license.\n4. Make sure the repository is ready for others to clone, install, and launch through Pinokio.\n5. Commit and push everything to GitHub, after confirming the repository's visibility and contents with me.\n6. Give me the published repository URL.\n\n\n**Include an install badge in the README.md.**\n\nAdd the following badge near the top of README.md, replacing OWNER/REPO with the actual GitHub repository:\n\n```\n[](https://pinokio.co/resolve?url=https://github.com/OWNER/REPO)\n```\n\nAfter publishing the GitHub repository, give the user the https://pinokio.co/resolve?url=https://github.com/OWNER/REPO install link and ask them to open it to add the project if prompted, signing in if needed.\n\n\n**Don't publish anything without my explicit approval.**\n\n\n**Start by interviewing me. Don't write code until we've established what we're building.**",
"intent": "create_app",
"fields": [],
"id": "01m76khmj5dzjffechznxdcr59",
"coverUrl": "https://api.pinokio.co/template-covers/01m76khmj5dzjffechznxdcr59/2497348f-c08e-4a1e-8f1a-f82a3b72e698.webp",
"publishedAt": "2026-11-12T01:34:14.217Z",
"unlistedAt": "2026-11-12T00:46:55.557Z",
"updatedAt": "2026-11-12T00:53:42.426Z",
"author": {
"username": "cocktailpeanut",
"name": "cocktailpeanut",
"avatarUrl": "https://assets.pinokio.co/avatars/optimized/1778687712237-wo1ws2elp6o.webp"
},
"substitutionRules": "Replace every original {{name}} occurrence in one pass. Whitespace around names is allowed. An odd backslash before {{...}} escapes a literal token; each preceding backslash pair emits one backslash. Other backslashes are unchanged. Never expand inserted answers or interpret replacement-string syntax. Defaults initialize answers; optional blanks are empty strings."
}