Modification of objects in film

1. A computer-implemented method of processing input video data comprising a sequence of image frames, the method comprising:

isolating an instance of an object within the sequence of image frames;

generating, using a machine learning model, a modified instance of the object;

determining optical flow data indicating an estimated warping relating the isolated instance of the object to the modified instance of the object over the sequence of image frames;

applying the estimated warping to the isolated instance of the object to generate a warped instance of the object;

adjusting colors of the modified instance of the object, based at least in part on the warped instance of the object, to generate a color-graded modified instance of the object; and

generating output video data in which at least part of the isolated instance of the object is replaced with a corresponding at least part of the color-graded modified instance of the object for at least some image frames of the sequence of image frames.

2. The computer-implemented method of claim 1 , wherein generating the output video data comprises:

determining parameter values for a synthetic model of the object, the first parameter values corresponding to the isolated instance of the object;

modifying the first parameter values for the synthetic model of the object to determine second parameter values for the synthetic model of the object, the second parameter values corresponding to the modified instance of the object;

incrementally interpolating between the first parameter values and the second parameter values over a subsequence of the sequence of image frames, thereby to determine interpolated parameter values for the synthetic model of the object;

generating, using the machine learning model, an incrementally modified instance of the object, based on the interpolated parameter values; and

replacing the isolated instance of the object with the incrementally modified instance of the object within the subsequence of the sequence of image frames.

3. The computer-implemented method of claim 1 , wherein generating the output video data comprises:

determining optical flow data indicating, for a subsequence of the sequence of image frames, an estimated warping relating the instance of the object to the modified instance of the object;

incrementally, over the subsequence of the sequence of image frames, applying the estimated warping to the instance of the object to determine an incrementally warped instance of the object;

incrementally, over the subsequence of the sequence of image frames, applying an inverse of the estimated warping to the modified instance of the object to determine an incrementally warped modified instance of the object; and

incrementally, over the subsequence of the sequence of image frames, dissolving the incrementally warped first instance of the object into the incrementally warped modified first instance of the object.

4. The computer-implemented method of claim 3 , wherein:

the incrementally dissolving takes place at a dissolving rate;

the incrementally applying the estimated warping and the inverse of the estimated warping takes place at a warping rate; and

a ratio of the dissolving rate to the warping rate increases to a maximum value then decreases within the subsequence of the sequence of image frames.

5. The computer-implemented method of claim 1 , further comprising:

blurring the warped instance of the object; and

blurring the modified instance of the object,

wherein adjusting the colors of the modified instance of the object is based on a pixelwise ratio of the blurred warped instance of the object and the blurred modified instance of the object, to generate the color-graded modified instance of the object.

6. The method of claim 5 , wherein each of the blurring of the warped instance of the object and the blurring of the modified instance of the object is performed using a blurring filter with a characteristic length scale of between 3 and 20 pixels.

7. The computer-implemented method of claim 1 , wherein the object is a human face.

8. The computer-implemented method of claim 7 , wherein the at least part of the isolated instance of the object includes a mouth but excludes eyes of the human face.

9. The computer-implemented method of claim 1 , further comprising:

detecting an event in the sequence of image frames and/or in an audio track associated with the sequence of image frames;

determining one or more image frames of the sequence of image frames at which the detected event takes place; and

determining a subsequence of the sequence of image frames in dependence on the determined one or more image frames at which the detected event takes place,

wherein generating the output video data comprises modifying the input video data to progressively transition, over the determined subsequence of the sequence of image frames, between the at least part of the isolated instance of the object and the corresponding at least part of the color-graded modified instance of the object.

10. The computer-implemented method of claim 9 , wherein determining the subsequence of the sequence of image frames is such that the subsequence of image frames ends before the event takes place.

11. The computer-implemented method of claim 9 , wherein:

the subsequence of the sequence of image frames is a first subsequence of the sequence of image frames; and

said modifying of the video data is to progressively transition from the at least part of the isolated instance of the object to the corresponding at least part of the modified instance of the object,

the method comprising further modifying the video data to progressively transition from the at least part of the modified instance of the object back to the corresponding at least part of the isolated instance of the object, over a second subsequence of the sequence of image frames.

12. The computer-implemented method of claim 9 , wherein the object is a human face, and the event is the human face uttering a plosive consonant or a bilabial nasal consonant.

13. The computer-implemented method of claim 1 , wherein generating the modified instance of the object comprises:

determining, using the isolated instance of the object, parameter values for a synthetic model of the object;

modifying the parameter values for the synthetic model of the object; and

rendering the modified instance of the object using the trained machine learning model and the modified parameter values for the synthetic model of the object.

14. The computer-implemented method of claim 13 , wherein the sequence of image frames is a first sequence of image frames, the instance of the object is a first instance of the object, and the parameter values for the instance of the object are second parameter values for the second instance of the object, the method further comprising:

identifying respective second instances of the object within a plurality of second sequences of image frames;

for at least some of the identified second instances of the object:

isolating said second instance of the object within image frames containing said instance of the object;

determining, using the isolated second instance of the object, associated second parameter values for the synthetic model of the object; and

training, using the isolated second instance of the object and the associated second parameter values for the synthetic model of the object, the machine learning model to reconstruct the isolated second instance of the object based at least in part on the associated second parameter values for the synthetic model of the object.

15. A non-transitory storage medium comprising machine readable instructions which, when executed by one or more processors, cause the one or more processors to perform a method of processing video data comprising a sequence of image frames, the method comprising:

isolating an instance of an object within the sequence of image frames;

generating, using a machine learning model, a modified instance of the object;

determining optical flow data indicating an estimated warping relating the isolated instance of the object to the modified instance of the object over the sequence of image frames;

applying the estimated warping to the isolated instance of the object to generate a warped instance of the object;

adjusting colors of the modified instance of the object, based at least in part on the warped instance of the object, to generate a color-graded modified instance of the object;

generating output video data in which at least part of the isolated instance of the object is replaced with a corresponding at least part of the color-graded modified instance of the object for at least some image frames of the sequence of image frames.

16. A system comprising at least one processor and at least one memory, wherein the memory comprises instructions to process input video data, the processing comprising:

isolating an instance of an object within the sequence of image frames;

generating, using a machine learning model, a modified instance of the object;

determining optical flow data indicating an estimated warping relating the isolated instance of the object to the modified instance of the object over the sequence of image frames;

applying the estimated warping to the isolated instance of the object to generate a warped instance of the object;

adjusting colors of the modified instance of the object, based at least in part on the warped instance of the object, to generate a color-graded modified instance of the object; and

generating output video data in which at least part of the isolated instance of the object is replaced with a corresponding at least part of the color-graded modified instance of the object for at least some image frames of the sequence of image frames.

17. The system of claim 16 , wherein generating the output video data comprises:

determining optical flow data indicating, for a subsequence of the sequence of image frames, an estimated warping relating the instance of the object to the modified instance of the object;

incrementally, over the subsequence of the sequence of image frames, applying the estimated warping to the instance of the object to determine an incrementally warped instance of the object;

incrementally, over the subsequence of the sequence of image frames, applying an inverse of the estimated warping to the modified instance of the object to determine an incrementally warped modified instance of the object; and

incrementally, over the subsequence of the sequence of image frames, dissolving the incrementally warped first instance of the object into the incrementally warped modified first instance of the object.

18. The system of claim 17 , wherein:

the incrementally dissolving takes place at a dissolving rate;

the incrementally applying the estimated warping and the inverse of the estimated warping takes place at a warping rate; and

a ratio of the dissolving rate to the warping rate increases to a maximum value then decreases within the subsequence of the sequence of image frames.

19. The system of claim 16 , wherein:

the object is a human face; and

the at least part of the isolated instance of the object includes a mouth but excludes eyes of the human face.