Midjourney Just Updated Its Most Useless Feature...
TLDRThe video script details a user's experience testing the new version of SL's image-to-prompt feature, comparing it to the old version. The user conducts a series of 10 challenges, uploading various images and evaluating the generated prompts' ability to recreate the images using Mid Journey Alpha. The results show improvement over the old version, with some images recreated well, while others, like the centaur, prove more difficult. The video concludes with a mention of a related video on creating consistent characters.
Takeaways
- 📷 The old version of SL had a feature for uploading an image and generating a prompt to recreate the image, but it was criticized for being too short and lacking detail.
- 🔄 The new version of SL is optimized for version six and is put to the test against 10 different challenges set by the user.
- 🖼️ Challenge one involved uploading an image and using SL's 'describe' feature to generate a prompt for Mid Journey Alpha to recreate the image.
- 🚀 The 'turbo mode' in Mid Journey Alpha was praised for its speed in generating images within 10 seconds.
- 🐾 The 'stone m' challenge, featuring a famous character, showed that SL describe sometimes struggled to capture the essence of the original image.
- 🌲 The 'gilberry' character challenge indicated that SL describe could do well with simpler characters but still had room for improvement.
- 🦉 The 'owl' challenge highlighted that SL describe could capture the subject's prominence but sometimes missed background details.
- 🌄 The 'magical sky' challenge tested SL describe's ability to incorporate background elements into the generated image.
- 🛹 The 'Roger on a skateboard in Time Square' challenge demonstrated the difficulty of capturing specific character and background details.
- 🐶 The 'dog' challenge showed that SL describe struggled with identifying and recreating specific dog breeds.
- 🤸♂️ The 'interesting pose' challenge revealed that SL describe could not accurately identify complex actions or poses.
- 🐎 The 'centaur' challenge was the most challenging, as SL describe failed to correctly identify and generate a centaur from the prompt.
Q & A
What was the main purpose of the old version of SL describe?
-The main purpose of the old version of SL describe was to allow users to upload an image and generate a prompt that could be used to recreate a new version of that image.
What were the limitations of the old version of SL describe according to the speaker?
-The limitations of the old version included short prompts that often lacked detail, reliance on keyword tags, and an inability to accurately maintain specific features like dog breeds or hairstyles from the original image.
How does the speaker describe the performance of the new version of SL describe?
-The speaker describes the new version of SL describe as optimized for version six and is testing it against 10 different challenges to see how it fares in comparison to the old version.
What was the outcome of the first challenge using the new SL describe?
-The first challenge resulted in a higher quality version of the original image, with the new SL describe generating a prompt that produced a similar image, although it missed some details like reflections on the face.
How does the Mid Journey Alpha platform factor into the testing process?
-Mid Journey Alpha is used as the platform where the generated prompts from SL describe are inputted to create new images in real-time, allowing for direct comparison with the original images.
What specific issue did the speaker encounter with the second challenge involving the stone m image?
-The second challenge resulted in images that did not closely resemble the original stone m image, with the AI struggling to capture the unique characteristics of the subject.
How did the new SL describe perform with the third challenge involving a gilberry character?
-The new SL describe performed well with the gilberry character, producing images that the speaker was not disappointed with, despite some differences from the original.
What was the most complicated character the speaker tested with SL describe?
-The most complicated character tested was a pure character, which the AI had difficulty accurately recreating based on the generated prompt.
What was the result of the challenge involving the centaur image?
-The challenge involving the centaur image was not successful, as the AI failed to correctly identify and generate a centaur, instead producing images of people riding horses.
What does the speaker suggest as an alternative to SL describe for creating consistent characters?
-The speaker suggests using image prompts in combination with SL describe as an alternative for creating consistent characters, as it can yield better results than using SL describe alone.
What was the speaker's overall conclusion about the new version of SL describe after the 10 challenges?
-The speaker concluded that while the new version of SL describe has improved capabilities, it still has limitations and struggles with certain challenges, such as accurately recreating specific characters and understanding complex concepts like a centaur.
Outlines
🔍 Evaluating SL Describe's Enhanced Prompt Generation
The script introduces the concept of using SL Describe for generating detailed prompts for recreating images through AI, comparing the new version optimized for version six against the old one, which lacked detail and often resulted in unsatisfactory recreations. The narrator sets up a challenge involving ten increasingly difficult tasks to test the capabilities of the new SL Describe, starting with simpler images and progressing to more complex scenes and characters. The first few challenges demonstrate improvements in image quality and detail, though not without minor issues in accurately capturing specific elements like reflections and textures.
🚀 Advanced Challenges Test SL Describe's Limits
As the video progresses, the challenges become more complex, testing SL Describe's ability to handle detailed backgrounds, character activities, and unique traits. Despite some successes, the tool struggles with specific tasks, such as accurately recreating a character on a skateboard in Times Square, maintaining consistency in character appearance across different prompts, and capturing the essence of a pet dog. The narrator highlights the tool's limitations when dealing with real-life images and more intricate compositions, illustrating the gap between AI-generated prompts and the nuanced requirements of certain images.
🎨 AI's Struggle with Highly Specific Imagery
The final segment delves into SL Describe's challenges with highly specific and imaginative prompts, such as recreating a centaur. Despite improvements in AI technology, the tool fails to accurately interpret and generate images for complex concepts that blend elements from different realms (like a human and horse hybrid). The script concludes with the narrator summarizing the experiment's findings, noting that while SL Describe shows promise in generating better-quality images than before, it still falls short in understanding and executing on very detailed and creative prompts.
Mindmap
Keywords
💡SL describe
💡Mid Journey Alpha
💡AI-generated images
💡Challenges
💡Prompts
💡Image quality
💡Character generation
💡Backgrounds
💡Consistent characters
💡Centaur
Highlights
The old version of SL's feature for uploading an image and using mid-journey to recreate it had limitations, such as short prompts and lack of detail.
The new version of SL is optimized for version six and is put to the test against 10 different challenges.
The first challenge involved uploading an image and using SL describe to generate a prompt for recreating the image.
SL describe's generated prompt was used in Mid Journey Alpha to recreate the image in real-time.
The second challenge featured a famous stone m image, which SL describe struggled to accurately recreate.
The third challenge involved a unique character, gilberry, which SL describe performed well on.
The fourth challenge was a complex character with a specific background, which SL describe handled adequately.
The fifth challenge required capturing the magic of the sky in the image, which SL describe managed to do.
The sixth challenge involved a character on a skateboard in Time Square, which SL describe accurately captured.
The seventh challenge tested SL describe's ability to create consistent characters, which showed a significant difference when using image prompts.
The eighth challenge aimed to recreate the user's dog, Jasmine, but SL describe failed to capture the breed.
The ninth challenge involved an AI-generated image of a person in a pose, which SL describe struggled to interpret correctly.
The final challenge was to create a centaur, a task that proved too difficult for SL describe.
The video concludes with a brief summary of the 10 challenges and a suggestion for a follow-up video on creating consistent characters.