Upload And Extend Audio · V4

Extend is the odd one out in the Suno endpoint family, and the most useful for anyone assembling a soundtrack in a graph. You give it a track and a timestamp, and it keeps going from there.

The practical case is the same one that plagues every video workflow: your generated clip is 90 seconds and the edit needs two minutes. Locally you'd be re-prompting and hoping; here you're asking the model to continue the actual audio you already have, which is a much better bet than generating a second track and crossfading it badly.

Filling the form

upload_url is required and is a URL string to the source track, capped at eight minutes by the tooltip. That's a URL field, so the usual pair is KIE • Upload Audio upstream - it takes a ComfyUI AUDIO object, converts it to 16-bit PCM WAV, hosts it, and hands you the string this node wants.

continue_at is the field that matters and the one people get wrong: the point in seconds where the extension begins, with the widget defaulting to 60. KIE's constraint is precise - greater than 0, less than the total duration of the uploaded audio. Point it past the end and you haven't extended anything, you've asked about a moment that doesn't exist.

prompt is the lyric text for the new section, used only if lyrics is empty, and - per the tooltip - it should not be passed at all when instrumental is on. Speaking of instrumental: the widget defaults to true while its own tooltip says it "defaults to false". The widget wins; if you want vocals, turn it off explicitly. That kind of drift between a provider's prose and its schema is normal in this ecosystem, and it's why reading the tooltips beats trusting them.

Then the same optional set as the other Suno nodes: style (200 characters on V4), title (100 characters here), negative_tags, vocal_gender (m or f, a nudge rather than a guarantee), style_weight, weirdness_constraint, audio_weight, lyrics, variety (0–4), persona_id with persona_model, and timeout_seconds, which at 1200 seconds is generous but not absurd for a music task. callback_url stays empty unless you're running a server that wants webhooks.

A note on the mode flag

Because ComfyUI sends every visible widget's value, leaving example lyric text in prompt while instrumental is on still ships that text to the provider. Clear fields you don't mean rather than relying on the mode switch to ignore them - the pack omits empty optional strings but doesn't second-guess a filled one.

Outputs

Seven, as with the rest of the Suno family: audio, url, all_urls_json, task_id, raw_json, credits_consumed and credits_left. The audio output is a native object, so wire it into a Save Audio or preview node, and if you're extending in stages, feed the saved result back through the upload node and extend again - that's how you build a long track out of a short one. Keep task_id in the graph if you're iterating; it's the only handle you have if you need to check on a job.

Install

cd ComfyUI/custom_nodes
git clone https://github.com/felipederosilva/ComfyUI-KIE-Nodes-Next
cd ComfyUI-KIE-Nodes-Next
pip install -r requirements.txt

Restart ComfyUI afterwards, then set your API key once under Settings → KIE.ai Nodes Next → Connection - the pack validates the key before saving it, shows a masked fingerprint, and confirms CONNECTED with your balance. ComfyUI Manager can install the pack for you (search "KIE.ai Nodes Next"), and on a headless box KIE_API_KEY in the environment takes priority over the saved key.

The version question

This is the V4 node, and KIE's docs list V4 as discontinued while the style limits in the tooltips (200 characters here versus 1000 on later versions) betray its age. It works and it's cheap. But extension quality matters more than most Suno features - you're asking the model to continue a musical idea coherently - and the newer model versions are meaningfully better at it. If the V6 variant of this node exists in your menu, use it, and treat this one as the compatibility path for a workflow you already built.