Gong Junmin
274 posts
Gong Junmin
@junmingong
engineer and researcher. @ACEStudio_en @acemusicAI
San Francisco, CA Katılım Ekim 2022
2.4K Takip Edilen903 Takipçiler
@junmingong It's less coherent than ACE-Step 1.5 musically, and doesn't have lyric input.
English
It’s amazing to see an open-source model surpass ACE-Step 1.5 so soon — smaller in size, faster in speed, and truly built for musicians. Good job!
Jordi Pons@jordiponsdotme
Stable Audio 3, explained in 5 figures. It’s a family of open-weight models for generating instrumental music and sound effects. The models are fast, support editing, and are trained on licensed and Creative Commons audio. 👾 artintech.substack.com/p/stable-audio… 🏋️♂️github.com/Stability-AI/s…
English
Runs on your machine.
· small runs on a MacBook Pro M4. It takes 5.92s to generate 120s of audio; with CoreML acceleration, this drops to 3s.
· medium requires ~6.5 GB VRAM for 120s of audio. It fits on consumer-grade GPUs and generates about 6m of audio in a few seconds.
English
Gong Junmin retweetledi
Stable Audio 3, explained in 5 figures.
It’s a family of open-weight models for generating instrumental music and sound effects.
The models are fast, support editing, and are trained on licensed and Creative Commons audio.
👾 artintech.substack.com/p/stable-audio…
🏋️♂️github.com/Stability-AI/s…
English
Khala 1.0 just dropped — a music generation model from the Central Conservatory of Music in Beijing. Paper, code, weights, and demo all open-sourced.
I gave a talk there recently on ACE-Step and got an early look at Khala. Excited to see it officially out. Open-source music gen is thriving.
💻 github.com/Khala-Music-AI…
📝 arxiv.org/abs/2605.01790
🎧 khala-music-ai.github.io/Khala-demo/
English
ACE‑Step 1.5 is a community‑owned model. It is the fastest‑growing open‑source music model and the best local music generation model for co‑building an ecosystem. I personally have drawn a great many inspiring ideas from it. Thank you all sincerely.
github.com/ace-step/ACE-S…
English
Gong Junmin retweetledi
なかなかいいものが作れたと思います！
音楽生成AIにて歌詞が破綻した部分をピンポイントに修正する方法を提案します！
生成AIなんでも展示会では、本内容や音楽生成AIの仕組みについてお話しします！
興味ある方お待ちしてます！
zenn.dev/asap/articles/… #zenn #生成AIなんでも展示会
日本語
It depends on your actual needs and aesthetic preferences. I don't think the model is incapable of producing this style; rather, there's a mismatch between how you describe it and how the model interprets it. Simply input "ambient", use the format tool to rewrite the caption to get a detailed description, then have Claude or ChatGPT reference that format to rewrite what you want.
English
@junmingong im trying to replace suno with ace step to make long (~4 min) ambient instrumental tracks but it gives me a hard time with excessively repetitive melodies and a total lack of bass frequencies no matter the prompt.
any tips? things got better using a bigger text encoder but still
English
Use ACE-Step 1.5-XL — own your model, don’t rent.
Doav@SodaNachos
@ElevenLabs for anyone interested: 1,800 credits per minute when creating an AI track. On the standard $5 plan that = approx. 7 songs at 3 minutes per song.
English
First of all, this model isn't for everyone. There's plenty of documentation and community tutorials on GitHub, and we're now in an era where coding isn't required at all. It has never been a barrier for those who are eager to explore.
The ComfyUI workflow has its limitations, which means many capabilities cannot be properly supported. The choice of workflow tool matters a lot. Even when I suggest new features to the ComfyUI team, getting them implemented is extremely difficult.
English
@junmingong I wish there was more documentation as to how to use Ace. I got it running in comfy, pumping songs out but when I want to cover something it’s lacklustre
English
Use ACE-Step 1.5-XL — own your model, don’t rent.
Brian Roemmele@BrianRoemmele
There are 20 companies that doing passable furniture music that is about 80% of most folks Spotify playlists. This tech is open source and local and free in 13 months. There is no moat for any music platforms. None. AI music with audience of 1.
English
@junmingong Is there a discord community that discusses fine tuning lora's for this model? I have hard drives full of break beats and drum samples that I want to train and share. A lot of it straight from vinyl recorded on tech 12's into pro-tools
English
I can't wait till people start claiming timestretched audio artifacting from early-mid 2000s music as being AI 😂
English
@entrepeneur4lyf Oh, my bad. I thought you were just another Suno fan here to attack open-source models and start arguments. Welcome to the ACE Step community.
English
@junmingong I do use ace step lol. I have songs published on acestep.app and didn’t the last 2 days making a desktop client.
English
Please don't say that. I am also a suno user. Suno and ACE-Step are different entities. People like to compare them, but they are fundamentally different. ACE-Step is not a model for everyone; it is more suitable for geeks who explore music experiments and requires a technical background, so it has a learning curve. Besides, their music generation preferences are also quite different. We recently updated to version 0.1.7, which offers better sound quality. I personally love it and even made a full album with it: The Space Between Notes discord.com/channels/13692…
How To AI@HowToAI_
Suno just lost its moat 🤯 Someone built ACE-Step UI, a Spotify-style interface for ACE-Step 1.5 that generates full songs with vocals up to 4+ minutes, completely locally on your own GPU. 100% Free. Open-Source and Local
English
@entrepeneur4lyf please check this one: discord.com/channels/13692…
English
@junmingong People say dumb stuff to get attention… I’d like to see him make this on ace step
English
Please don't say that. I am also a suno user. Suno and ACE-Step are different entities. People like to compare them, but they are fundamentally different. ACE-Step is not a model for everyone; it is more suitable for geeks who explore music experiments and requires a technical background, so it has a learning curve. Besides, their music generation preferences are also quite different. We recently updated to version 0.1.7, which offers better sound quality. I personally love it and even made a full album with it: The Space Between Notes discord.com/channels/13692…
English
