Assigned
Status Update
No update yet.
Comments
wr...@gmail.com
Hello,
Let me clarify and further elaborate on the issues experienced with Imagen
3, highlighting technical shortcomings and potential improvements that
could significantly enhance user experience and competitive positioning.
Firstly, the most pressing issue is the quality of images generated by
Imagen 3. Approximately 70% of the images produced are unacceptable in
terms of clarity and detail. I've already provided detailed feedback on
this previously, but unfortunately, due to a lack of proper dialogue and
response, it feels like my feedback simply disappears into a black hole.
In practical terms, whether using Vertex AI or via browser-based access,
the generated images frequently appear severely grainy, distorted, and
lacking crucial details. For instance, even when paying extra for
high-resolution upscaling through Vertex AI (up to 5K), the result often
resembles a poorly rendered 240p image. There clearly isn't an effective,
high-end upscaling technology employed here, such as AI-driven
super-resolution, GAN-based upscaling methods, or diffusion-based
refinement, all of which could drastically improve the final image output.
Another critical issue involves pricing and resource utilization.
Freepik.com charges 50 credits per image theoretically, but in reality, it
deducts around 100 credits, severely overpricing the product compared to
alternative market solutions. To illustrate, Imagen 3 generates an image
within 3-5 seconds, utilizing roughly 12-16 compute units. Conversely, FLUX
1.1 PRO ULTRA, though requiring 160-175 compute units over approximately 30
seconds, delivers vastly superior results in terms of clarity, detail
preservation, and overall quality, allowing effective 2-3x scaling without
distortion. Attached, you'll find examples of theoretical 4K upscaling from
Imagen 3, which visually confirm its inadequacies and highlight that FLUX
1.1 PRO ULTRA delivers significantly more value at a lower effective cost.
It's acknowledged that Freepik.com's pricing structure is not solely
determined by Google. However, it clearly indicates Imagen 3's current
inability to compete effectively at this pricing level. While mobile users
might find image quality acceptable due to smaller screen sizes,
professional users who demand higher standards will inevitably be
disappointed when viewing images on larger or higher-resolution screens.
Moreover, the current model seems primarily designed with overly
restrictive guidelines, creating issues such as context misinterpretation,
erroneous child protection interventions, and content generation
hallucinations even without explicit problematic keywords in prompts. These
restrictions severely undermine creative freedom and social media
compatibility, mixing concepts of intimacy with explicit adult content
incorrectly, thus limiting the scope of practical usage.
From an ethical and strategic standpoint, it's problematic that meaningful
support interactions, such as discussing critical flaws, require
prohibitively expensive upfront payments (e.g., $15,000). This creates
significant ethical, legal, and user trust concerns, effectively silencing
valuable user feedback and hindering effective communication. If widely
publicized, it could critically damage user trust.
There also seems to be a serious internal communication gap within Google
regarding feedback management. When feedback is consistently ignored or
superficially addressed, critical issues remain unresolved. This suggests
either internal miscommunication or deliberate avoidance, resulting in
products being marketed prematurely as revolutionary innovations without
genuinely addressing underlying issues.
Regarding VEO 2 specifically, I reviewed the provided documentation, and it
felt extremely basic—akin to a children's creativity guide rather than
professional documentation. Additionally, Freepik charges 1000 credits for
a 5-second video that takes considerable time to generate, whereas Kling AI
generates full HD videos in less than one second, offering vastly superior
cost-efficiency. Currently, VEO 2 is approximately 30 times more expensive
and notably slower. Even compared to the WAN 2.1 model running on MimicPC
(costing around $1.25 per hour on a 48GB VRAM server), VEO 2 falls behind
in both performance and economic feasibility.
My proposal to resolve these challenges is straightforward: Introduce an
optional Gemini AI-linked subscription model, priced attractively around
$35, offering approximately 250 images and 10-20 minutes of video
generation per month. Clearly indicating the experimental nature of the
product during beta phases, coupled with minimalistic, non-intrusive
watermarks, could significantly increase user engagement and provide
valuable revenue streams during development. OpenAI's pricing ($200) is
significantly higher, and their quality doesn't justify such a price.
Google could capitalize on this competitive gap, clearly highlighting
superior quality at substantially lower prices, making it attractive for
creators and social media professionals alike.
Lastly, immediate access to VEO 2 would allow me to further elaborate on
advanced features and improvements, potentially transforming Google's AI
products into unmatched industry leaders. I'm more than ready to offer
detailed strategic insights and practical suggestions to solidify Google's
position in the AI innovation race, creating genuinely desirable and highly
competitive products.
Thank you for your attention—I hope we can collaborate effectively to
enhance Imagen 3 and VEO 2 significantly.
... I kindly request access to VEO 2 via Vertex AI and web-based platforms
for further testing and evaluation purposes.
<buganizer-system@google.com> ezt írta (időpont: 2025. márc. 6., Cs, 7:55):
Let me clarify and further elaborate on the issues experienced with Imagen
3, highlighting technical shortcomings and potential improvements that
could significantly enhance user experience and competitive positioning.
Firstly, the most pressing issue is the quality of images generated by
Imagen 3. Approximately 70% of the images produced are unacceptable in
terms of clarity and detail. I've already provided detailed feedback on
this previously, but unfortunately, due to a lack of proper dialogue and
response, it feels like my feedback simply disappears into a black hole.
In practical terms, whether using Vertex AI or via browser-based access,
the generated images frequently appear severely grainy, distorted, and
lacking crucial details. For instance, even when paying extra for
high-resolution upscaling through Vertex AI (up to 5K), the result often
resembles a poorly rendered 240p image. There clearly isn't an effective,
high-end upscaling technology employed here, such as AI-driven
super-resolution, GAN-based upscaling methods, or diffusion-based
refinement, all of which could drastically improve the final image output.
Another critical issue involves pricing and resource utilization.
Freepik.com charges 50 credits per image theoretically, but in reality, it
deducts around 100 credits, severely overpricing the product compared to
alternative market solutions. To illustrate, Imagen 3 generates an image
within 3-5 seconds, utilizing roughly 12-16 compute units. Conversely, FLUX
1.1 PRO ULTRA, though requiring 160-175 compute units over approximately 30
seconds, delivers vastly superior results in terms of clarity, detail
preservation, and overall quality, allowing effective 2-3x scaling without
distortion. Attached, you'll find examples of theoretical 4K upscaling from
Imagen 3, which visually confirm its inadequacies and highlight that FLUX
1.1 PRO ULTRA delivers significantly more value at a lower effective cost.
It's acknowledged that Freepik.com's pricing structure is not solely
determined by Google. However, it clearly indicates Imagen 3's current
inability to compete effectively at this pricing level. While mobile users
might find image quality acceptable due to smaller screen sizes,
professional users who demand higher standards will inevitably be
disappointed when viewing images on larger or higher-resolution screens.
Moreover, the current model seems primarily designed with overly
restrictive guidelines, creating issues such as context misinterpretation,
erroneous child protection interventions, and content generation
hallucinations even without explicit problematic keywords in prompts. These
restrictions severely undermine creative freedom and social media
compatibility, mixing concepts of intimacy with explicit adult content
incorrectly, thus limiting the scope of practical usage.
From an ethical and strategic standpoint, it's problematic that meaningful
support interactions, such as discussing critical flaws, require
prohibitively expensive upfront payments (e.g., $15,000). This creates
significant ethical, legal, and user trust concerns, effectively silencing
valuable user feedback and hindering effective communication. If widely
publicized, it could critically damage user trust.
There also seems to be a serious internal communication gap within Google
regarding feedback management. When feedback is consistently ignored or
superficially addressed, critical issues remain unresolved. This suggests
either internal miscommunication or deliberate avoidance, resulting in
products being marketed prematurely as revolutionary innovations without
genuinely addressing underlying issues.
Regarding VEO 2 specifically, I reviewed the provided documentation, and it
felt extremely basic—akin to a children's creativity guide rather than
professional documentation. Additionally, Freepik charges 1000 credits for
a 5-second video that takes considerable time to generate, whereas Kling AI
generates full HD videos in less than one second, offering vastly superior
cost-efficiency. Currently, VEO 2 is approximately 30 times more expensive
and notably slower. Even compared to the WAN 2.1 model running on MimicPC
(costing around $1.25 per hour on a 48GB VRAM server), VEO 2 falls behind
in both performance and economic feasibility.
My proposal to resolve these challenges is straightforward: Introduce an
optional Gemini AI-linked subscription model, priced attractively around
$35, offering approximately 250 images and 10-20 minutes of video
generation per month. Clearly indicating the experimental nature of the
product during beta phases, coupled with minimalistic, non-intrusive
watermarks, could significantly increase user engagement and provide
valuable revenue streams during development. OpenAI's pricing ($200) is
significantly higher, and their quality doesn't justify such a price.
Google could capitalize on this competitive gap, clearly highlighting
superior quality at substantially lower prices, making it attractive for
creators and social media professionals alike.
Lastly, immediate access to VEO 2 would allow me to further elaborate on
advanced features and improvements, potentially transforming Google's AI
products into unmatched industry leaders. I'm more than ready to offer
detailed strategic insights and practical suggestions to solidify Google's
position in the AI innovation race, creating genuinely desirable and highly
competitive products.
Thank you for your attention—I hope we can collaborate effectively to
enhance Imagen 3 and VEO 2 significantly.
... I kindly request access to VEO 2 via Vertex AI and web-based platforms
for further testing and evaluation purposes.
<buganizer-system@google.com> ezt írta (időpont: 2025. márc. 6., Cs, 7:55):
va...@google.com
ja...@google.com
Hello,
Thank you for reaching out to us with your request.
We have duly noted your feedback and will thoroughly validate it. While we cannot provide an estimated time of implementation or guarantee the fulfillment of the issue, please be assured that your input is highly valued. Your feedback enables us to enhance our products and services.
We appreciate your continued trust and support in improving our Google Cloud Platform products. In case you want to report a new issue, Please do not hesitate to create a new issue on the
Once again, we sincerely appreciate your valuable feedback; Thank you for your understanding and collaboration.
ja...@google.com
Additionally, I would like to request access to Veo 2. I am working on
building a new community, encouraging more people from my current community
to move towards the artistic direction. I want to use Veo 2 to excite and
engage my new community. This also includes promoting Veo 2 on external
platforms, increasing discussions about it and making the model even more
sought-after, while also reaching and activating a new audience.
<buganizer-system@google.com> ezt írta (időpont: 2025. márc. 13., Cs, 7:20):
building a new community, encouraging more people from my current community
to move towards the artistic direction. I want to use Veo 2 to excite and
engage my new community. This also includes promoting Veo 2 on external
platforms, increasing discussions about it and making the model even more
sought-after, while also reaching and activating a new audience.
<buganizer-system@google.com> ezt írta (időpont: 2025. márc. 13., Cs, 7:20):
Description
Issue summary: The customer is facing “OOM and OOD” errors for the jobs and are being retried. And also the customer wants metrics that show disk usage.