YouTube to Add Prompt-Generated Custom Feeds, GIF Comments and Live Dubbing
YouTube will let users in the US generate custom video feeds from conversational prompts across web, mobile and TV.
The option will live on the homepage. Users select "Your custom feed" from the options at the top of the Home tab and enter a prompt describing the type of feed they want, according to YouTube support documentation. The feed then generates as a separate space, accessible from a tab on the homepage. Users in the US will be able to try the system soon on all three surfaces, Engadget reported on Sept. 23.
The scope has widened. YouTube has expanded the custom feeds option to enable creation of multiple personalized feeds rather than a single filtered view, SocialMediaToday reported in August. Users can edit the prompt at any time in the text box at the top of the custom feed to generate a new space, 9to5Google reported in May. Discovery becomes a prompt.
Comments get GIFs and synthetic voice
YouTube will soon allow users to leave a GIF under a video via a searchable GIF library. GIF comments will be available in the mobile app globally sometime this fall.
The TV app is taking a different path. It now lets users post voice comments via a text-to-speech tool where the TV supports it. That addresses a practical input constraint. Typing with a remote remains slow, and voice with synthesis lowers the cost of participation from the couch without requiring a paired phone.
For a tech-literate reader, the moderation question is immediate. Text comment classifiers, GIF libraries and synthetic speech have different abuse surfaces. A searchable library implies curation and rate limits rather than open upload, which keeps the problem bounded.
Assistants, shows and live translation
YouTube is expanding Ask YouTube to include shopping features such as video recommendations and a comparison table for product review requests. The change moves the assistant from summarization toward structured commerce retrieval. A user asking for product reviews gets not only videos but a normalized comparison.
YouTube Music's Ask Music tool will let users learn more about an artist and adjust their song queue. That pairs knowledge retrieval with queue management in one conversational loop.
YouTube is also introducing Podcast Lineup, a genAI-powered feature offering a short weekly preview of suggested podcasts. The format is a digest, not a full transcript or auto-playlist. It is designed for sampling.
On the creator side, YouTube expanded its shows feature to Shorts, letting creators arrange Shorts episodically with custom thumbnails and sequential playback. That brings series metadata, thumbnails and ordering, conventions long associated with episodic television, to vertical short form.
YouTube will start piloting a Live auto-dubbing tool in early 2027, initially for English and Spanish. Live dubbing has tighter constraints than offline dubbing. Latency, speaker diarization and translation quality must hold up under real-time delivery. Starting with two high-resource languages limits the variable space.
Separately, YouTube extended its exclusive Coachella streaming rights through 2030.
The broader context here is familiar to anyone who has operated large-scale recommendation. The Home feed optimizes for aggregate retention. Custom feeds push control to the user and fragment that optimization problem into many smaller, prompt-defined surfaces. That can improve satisfaction for intent-driven viewing, such as learning a skill or following a niche, while making measurement harder. Success is no longer one ranking to tune.
In my view, the trade to watch is personalization against shared context. I raised two children through the PC era into mobile and streaming, and the pattern was consistent. The more control they got over their feeds, the faster they found what they wanted and the narrower what they encountered became. Prompt-generated feeds will likely do the same for YouTube viewers. The tooling is genuinely useful. It also puts more responsibility on the user to seek breadth.
Looking at what this means for builders, the near-term work is unglamorous. Prompt parsing for feed generation, GIF library governance, text-to-speech reliability across TV hardware, and low-latency speech translation for live streams all need to work at YouTube scale before the features matter. If they do, what this enables is straightforward. Viewers get more precise control over discovery, creators get new packaging for Shorts and live audiences, and conversational assistants handle more of the navigation between them.
