Choose by the image’s job in the sequence. Use LTXV Add Guide when an image or video should guide a specified position in a latent sequence. Use LTXV ImgToVideo Inplace when the video latent already exists and the still should condition its opening frames. Use LTXV ImgToVideo when starting a new video latent from a still. These are different insertion stages, not interchangeable names for the same operation.
Decide and connect the path
- Write down the intended timeline role: opening appearance for the whole clip, a reference at a particular point, or the starting image for a newly created sequence. Treat these as project requirements, not node guarantees about the final visual result.
- Check whether you already have the target video latent. If not, and the still should start the video, choose LTXV ImgToVideo. It creates a video-sized latent with the encoded image at its beginning. If the latent exists, continue to the next decision.
- For an existing latent, choose Inplace when the image belongs at the beginning. Choose Add Guide when guidance belongs at a specified frame position; connect the existing latent and conditioning inputs it requires. In the graph, confirm sockets by their type labels, not color alone.
- Set strength according to the chosen node’s documented meaning. Inplace strength controls conditioning of initial frames, and its documentation also describes the initial-frame noise mask. Add Guide strength controls conditioning influence. ImgToVideo strength describes preservation versus modification of the original image content in the generated video’s first frames. Do not assume the three controls have interchangeable scales or visual effects.
- For Add Guide, check that the guide fits the latent at the selected position. Its image/video input must follow the 8*n+1 frame pattern; an invalid video length may be cropped to a nearby valid length. For long guides, the documented frame-index alignment rule also matters. See the Add Guide node documentation before choosing that branch.
Hypothetical acceptance worksheet
The table below uses assumed project observations to help distinguish the paths. It is a proposed review method, not a report of software testing or guaranteed output.
| Assumed request and observation | Latent before image node | Decision and acceptance check |
|---|---|---|
| A 97-frame clip should begin from a product still; the graph has no video latent yet. | No | Choose ImgToVideo; confirm the new latent has the requested video length and the image is assigned to its beginning. |
| A 97-frame latent already exists; the still is intended to set the opening appearance. | Yes | Choose Inplace; review the generated opening against the brief and record the selected strength. |
| An existing sequence should use a 9-frame reference starting at frame 24. | Yes | Choose Add Guide; 9 follows 8*n+1, and frame 24 is divisible by 8. Check that the guide fits the remaining sequence. |
For the last row, these values are hypothetical and illustrate documented constraints, not a claim that the resulting scene will match the reference. Keep the acceptance question focused: did the graph place the image in the intended temporal role, and does the rendered result meet the project brief?
Limits and next steps
The official descriptions do not promise an exact degree of visual similarity, identity retention, or motion behavior. Avoid translating a strength value into a percentage of pixels preserved. Inplace documents a 0–1 range; ImgToVideo also documents 0–1; Add Guide lists a different range. Follow the selected node’s own input range.
If Add Guide rejects the graph, check guide length, frame position, and whether all guide frames fit the latent. Its documentation also says adding a guide to a latent combining audio and video channels is unsupported. If the wrong part of the clip is conditioned, reconsider the path before adjusting strength: an opening condition is not a substitute for a guide positioned later. For ImgToVideo, check the requested dimensions and length; its image is resized and center-cropped to the requested dimensions, using RGB channels. For Inplace, the image is resized for the VAE and only RGB is used. These documented preprocessing details may affect what reaches the latent; they do not establish the appearance of the final generation. See the ImgToVideo Inplace documentation and ImgToVideo documentation.
Three quick questions
Can I use Add Guide to start a new video from one image? Add Guide modifies a supplied latent and conditioning. If the target latent does not exist and the image should start a new sequence, ImgToVideo is the documented route.
Does Inplace place the image at any chosen frame? Its documentation specifies conditioning the initial frames of an existing latent, not selecting an arbitrary frame position. Use Add Guide for a specified position.
Does strength 1 guarantee an identical opening frame? No such guarantee is stated for the final generated video. Read the selected node’s documented conditioning or preservation meaning, then judge the output against the project’s own acceptance criteria.
Sources and next step
- LTXVAddGuide - ComfyUI Built-in Node Documentation
- LTXVImgToVideoInplace - ComfyUI Built-in Node Documentation
- LTXVImgToVideo - ComfyUI Built-in Node Documentation
This guide interprets public documentation and uses explicitly hypothetical acceptance examples, not application tests. Controls and availability can change; verify the cited manual against your installed version.
For work created in Imagild, carry the input requirements and acceptance checklist into the film workspace. The external software controls discussed above belong to their respective products.