@monokee here is the url

https://app.color.io/?preset=%28%27o%2540oNi%2540iNs*%253CQ9.%257F22%252C1.4868F53.4815%257B564%2524-7661%2525-7845q6%253E8%25204914%255DjQ%253E.2772%257B753F%257F.7856%253A539%2524Y175%2525Y74q3923%25203982%255D%2526separa%252B%253B29%7E%257CVolumeZ6146%257B057%2526%257CBalanceZ6306Y17%2526%253CsZ3469-2864jsZ5025%253A975%2560d*hueI4744W167YW3597Y5%253EW%257F94Y892W8089Y249%2522%253A744%255D%2526chro%253FI5W5994%257B481%2522-%257D83%255D%2526lu%253FI4%257D9W3269%257B05%2522%253A034%255D%2560l*lu%253Fk0553W2%257F7-1683W4859Y026%2522-9315%255D%2526huek46%253EW1505Y783W3563%253A142W6089-2662W8008-3222%2522%253A6%253E%255D%2560r*%2523%252F%253B%7E%2523Chro%253F%253B%7E%2523%253D%7Esoftness%253B%7Eacutance%210%7Edistribu%252BU1JMix%253BJRadius%253BJThresholdU5JHueSatZ5Y%255DKExposureU5K%253DKRoundnessU5KCenterZ5Y%255D%29%29*%2540bypass%21false%7EmixAmount%21100%7E-%252CGF%255D%252C%255BG0.IVs%252F%255E0-J%7Ehala%252BK%7EvignetteNdt%21%27EXPORT_ACES_CCT%27%29%7EQMapVectors%255EU%21GWFGY-5Z%21%255BGj%2526highlightkVsLu%253F%255E0-qF2%257D.51%257D-%2520F298.0882-%2522F1%2523grain%2524%252264.4934%2525%252292.9429%2526%255D%7E%252Btion%252FDensity%253A-4%253BU3%253Cshadow%253DSizeU5%253E18%253Fma%2540%21%28%27%255E%21%255B%255B%2560%255D%29%7E%257B-6%257Ccolor%257D33%257F58%2501%257F%257D%257C%257B%2560%255E%2540%253F%253E%253D%253C%253B%253A%252F%252B%2526%2525%2524%2523%2522%2520qkjZYWUQNKJIGF-*&

@guillaume I just double checked everything in the implementation to make sure the color space transforms are correct. Color.io uses ACES 1.3 and the data from this white paper: https://s3.amazonaws.com/red_3/downloads/other/white-papers/REDWIDEGAMUTRGB%20AND%20LOG3G10%20Rev-B.pdf for the color space transform. I can confirm that the conversion from Log3G10 to Rec709 in Color.io is correct and will produce exactly the curve you posted. You will get the same results when you generate the same color space transform with Davinci Resolve or other tools. The "hard clipping" in the curve visualization you mentioned does not really tell you anything about how the curve will transform an actual camera signal. It will not be a real issue with actual Log3G10 footage because the signal requires this conversion in order to look correct.