🟠 Mach8 | Midjourney is surpassed big time
The launch of Flux.1 by Black Forest Labs, an open-source challenger redefining image generation, the latest in AI voice technology and automated meeting solutions, we bring you the cutting-edge.
High-Quality Open-Source Image Generation
A German-based Gen-AI research company called Black Forest Labs has just launched Flux.1. It is, by far, the highest quality open-source image generation model to date. In fact, it even rivals, or perhaps surpasses, the image quality of Midjourney.
(Images generated by Flux.1 Link to image)
Why Open-Source Matters for Text-to-Image Generation
The main advantage of Flux.1 is that it’s fully open-source and available for anyone to download and use. This opens the door to a wide range of new applications that were previously unthinkable.
One key benefit is the ability to fine-tune custom image models with LoRA. We can create our own custom LoRAs to fine-tune Flux models, allowing us to upload a specific set of assets that the model can use to better represent logos, colors, and compositions.
Another advantage is the cost-effectiveness of generating images. Instead of paying a company like Midjourney for model usage plus a margin, you now only need to cover computing costs, significantly reducing the overall expense of running image generation models.
Additionally, there’s no vendor lock-in, and you have full control over API capabilities. You won’t have to worry about increasing rates, poor support, or even being blocked from the service. By hosting the model yourself, you can build your own APIs on top of it.
Try It Out Yourself
It’s very easy to give it a try. Simply go to Replicate and open the flux-dev model. Enter your prompt and click "Run." You can also download the model from GitHub and run it locally on your own machine.
GPT Advanced Voice Mode and Gemini Live
With ChatGPT’s Advanced Voice Mode gradually rolling out and Google launching Gemini Live, the AI voice showdown has begun. Are these new tools as limited as Siri, or are they the advanced AI assistants we've been promised?
GPT Mimicking Users' Voices
The initial online reactions to this new tool are overwhelmingly positive. It truly feels like you're conversing with a human rather than a computer. Watch a example here
But as you can see in this demo, it’s also allegedly mimicking users' voices. While this capability is impressive, it’s also quite unsettling. It indicates that this is not just a fast transcriber or voice cloner but a full-fledged audio model.
Google’s Alternative Launched
Following OpenAI’s lead, Google has also launched its new AI voice companion, now available on all new Pixel phones. But how does it compare to OpenAI’s Advanced Voice Mode?
According to reviews and first impressions, Google’s offering performs well, though with slightly slower response times and less compelling voice quality compared to OpenAI’s alternative. However, features like interrupting and fluent conversation seem to work very well. Watch a preview here.
Automated Meeting Notes with Cockatoo
Many of us have calendars filled with meetings, leading to a flood of notes and a lot of time spent rewriting them into clear action lists and summaries, especially for half-year review meetings. However, there are methods to improve your efficiency in meeting documentation. We’ve created a simple workflow to automate this process—just follow these steps:
Create an Account on Cockatoo, Start by creating an account on Cockatoo, which allows you to transcribe all your meetings on the fly. Visit their website to sign up.
Export Your Transcripts, The transcripts can be lengthy and messy, so copy them into ChatGPT with a prompt like this:
“Below is a transcript of a half-year review meeting I just had. Can you create a summary document of all the topics we discussed? Please format it nicely and make it easily readable.”
Paste your transcript after the prompt.Copy to a Google Doc, Once ChatGPT has processed the transcript, copy the output into a Google Doc, and you’re done! You now have a fully finished meeting document.